EFLAGS)))]>;
}
+//===----------------------------------------------------------------------===//
+// SSE 1 & 2 Instructions Classes
+//===----------------------------------------------------------------------===//
+
+/// sse12_fp_scalar - SSE 1 & 2 scalar instructions class
+multiclass sse12_fp_scalar<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ RegisterClass RC, X86MemOperand x86memop> {
+ let isCommutable = 1 in {
+ def rr : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
+ OpcodeStr, [(set RC:$dst, (OpNode RC:$src1, RC:$src2))]>;
+ }
+ def rm : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
+ OpcodeStr, [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))]>;
+}
+
+/// sse12_fp_scalar_int - SSE 1 & 2 scalar instructions intrinsics class
+multiclass sse12_fp_scalar_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
+ string asm, string SSEVer, string FPSizeStr,
+ Operand memopr, ComplexPattern mem_cpat> {
+ def rr_Int : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
+ asm, [(set RC:$dst, (
+ !nameconcat<Intrinsic>("int_x86_sse",
+ !strconcat(SSEVer, !strconcat("_",
+ !strconcat(OpcodeStr, FPSizeStr))))
+ RC:$src1, RC:$src2))]>;
+ def rm_Int : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memopr:$src2),
+ asm, [(set RC:$dst, (
+ !nameconcat<Intrinsic>("int_x86_sse",
+ !strconcat(SSEVer, !strconcat("_",
+ !strconcat(OpcodeStr, FPSizeStr))))
+ RC:$src1, mem_cpat:$src2))]>;
+}
+
+/// sse12_fp_packed - SSE 1 & 2 packed instructions class
+multiclass sse12_fp_packed<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ RegisterClass RC, ValueType vt,
+ X86MemOperand x86memop, PatFrag mem_frag,
+ Domain d, bit MayLoad = 0> {
+ let isCommutable = 1 in
+ def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
+ OpcodeStr, [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))],d>;
+ let mayLoad = MayLoad in
+ def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
+ OpcodeStr, [(set RC:$dst, (OpNode RC:$src1,
+ (mem_frag addr:$src2)))],d>;
+}
+
+/// sse12_fp_packed_logical_rm - SSE 1 & 2 packed instructions class
+multiclass sse12_fp_packed_logical_rm<bits<8> opc, RegisterClass RC, Domain d,
+ string OpcodeStr, X86MemOperand x86memop,
+ list<dag> pat_rr, list<dag> pat_rm> {
+ let isCommutable = 1 in
+ def rr : PI<opc, MRMSrcReg, (outs RC:$dst),
+ (ins RC:$src1, RC:$src2), OpcodeStr, pat_rr, d>;
+ def rm : PI<opc, MRMSrcMem, (outs RC:$dst),
+ (ins RC:$src1, x86memop:$src2), OpcodeStr, pat_rm, d>;
+}
+
+/// sse12_fp_packed_int - SSE 1 & 2 packed instructions intrinsics class
+multiclass sse12_fp_packed_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
+ string asm, string SSEVer, string FPSizeStr,
+ X86MemOperand x86memop, PatFrag mem_frag,
+ Domain d> {
+ def rr_Int : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
+ asm, [(set RC:$dst, (
+ !nameconcat<Intrinsic>("int_x86_sse",
+ !strconcat(SSEVer, !strconcat("_",
+ !strconcat(OpcodeStr, FPSizeStr))))
+ RC:$src1, RC:$src2))], d>;
+ def rm_Int : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
+ asm, [(set RC:$dst, (
+ !nameconcat<Intrinsic>("int_x86_sse",
+ !strconcat(SSEVer, !strconcat("_",
+ !strconcat(OpcodeStr, FPSizeStr))))
+ RC:$src1, (mem_frag addr:$src2)))], d>;
+}
+
//===----------------------------------------------------------------------===//
// SSE1 Instructions
//===----------------------------------------------------------------------===//
-// Move Instructions. Register-to-register movss is not used for FR32
-// register copies because it's a partial register update; FsMOVAPSrr is
-// used instead. Register-to-register movss is not modeled as an INSERT_SUBREG
-// because INSERT_SUBREG requires that the insert be implementable in terms of
-// a copy, and just mentioned, we don't use movss for copies.
-let Constraints = "$src1 = $dst" in
+// Conversion Instructions
+
+// Match intrinsics which expect XMM operand(s).
+def CVTSS2SIrr: SSI<0x2D, MRMSrcReg, (outs GR32:$dst), (ins FR32:$src),
+ "cvtss2si{l}\t{$src, $dst|$dst, $src}", []>;
+def CVTSS2SIrm: SSI<0x2D, MRMSrcMem, (outs GR32:$dst), (ins f32mem:$src),
+ "cvtss2si{l}\t{$src, $dst|$dst, $src}", []>;
+
+def CVTDQ2PSrr : PSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PSrm : PSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
+
+// Aliases for intrinsics
+def Int_CVTTSS2SIrr : SSI<0x2C, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
+ "cvttss2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst,
+ (int_x86_sse_cvttss2si VR128:$src))]>;
+def Int_CVTTSS2SIrm : SSI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f32mem:$src),
+ "cvttss2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst,
+ (int_x86_sse_cvttss2si(load addr:$src)))]>;
+
+let Constraints = "$src1 = $dst" in {
+ def Int_CVTSI2SSrr : SSI<0x2A, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, GR32:$src2),
+ "cvtsi2ss\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst, (int_x86_sse_cvtsi2ss VR128:$src1,
+ GR32:$src2))]>;
+ def Int_CVTSI2SSrm : SSI<0x2A, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, i32mem:$src2),
+ "cvtsi2ss\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst, (int_x86_sse_cvtsi2ss VR128:$src1,
+ (loadi32 addr:$src2)))]>;
+}
+
+// Compare Instructions
+let Defs = [EFLAGS] in {
+def COMISSrr: PSI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
+ "comiss\t{$src2, $src1|$src1, $src2}", []>;
+def COMISSrm: PSI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
+ "comiss\t{$src2, $src1|$src1, $src2}", []>;
+} // Defs = [EFLAGS]
+
+//===----------------------------------------------------------------------===//
+// SSE 1 & 2 - Move Instructions
+//===----------------------------------------------------------------------===//
+
+// Move Instructions. Register-to-register movss/movsd is not used for FR32/64
+// register copies because it's a partial register update; FsMOVAPSrr/FsMOVAPDrr
+// is used instead. Register-to-register movss/movsd is not modeled as an
+// INSERT_SUBREG because INSERT_SUBREG requires that the insert be implementable
+// in terms of a copy, and just mentioned, we don't use movss/movsd for copies.
+let Constraints = "$src1 = $dst" in {
def MOVSSrr : SSI<0x10, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, FR32:$src2),
"movss\t{$src2, $dst|$dst, $src2}",
[(set (v4f32 VR128:$dst),
(movl VR128:$src1, (scalar_to_vector FR32:$src2)))]>;
+def MOVSDrr : SDI<0x10, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, FR64:$src2),
+ "movsd\t{$src2, $dst|$dst, $src2}",
+ [(set (v2f64 VR128:$dst),
+ (movl VR128:$src1, (scalar_to_vector FR64:$src2)))]>;
+}
+// Loading from memory automatically zeroing upper bits.
+let canFoldAsLoad = 1, isReMaterializable = 1 in {
+def MOVSSrm : SSI<0x10, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
+ "movss\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (loadf32 addr:$src))]>;
+let AddedComplexity = 20 in
+def MOVSDrm : SDI<0x10, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
+ "movsd\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (loadf64 addr:$src))]>;
+}
+
+let AddedComplexity = 15 in {
// Extract the low 32-bit value from one vector and insert it into another.
-let AddedComplexity = 15 in
def : Pat<(v4f32 (movl VR128:$src1, VR128:$src2)),
(MOVSSrr (v4f32 VR128:$src1),
(EXTRACT_SUBREG (v4f32 VR128:$src2), sub_ss))>;
+// Extract the low 64-bit value from one vector and insert it into another.
+def : Pat<(v2f64 (movl VR128:$src1, VR128:$src2)),
+ (MOVSDrr (v2f64 VR128:$src1),
+ (EXTRACT_SUBREG (v2f64 VR128:$src2), sub_sd))>;
+}
// Implicitly promote a 32-bit scalar to a vector.
def : Pat<(v4f32 (scalar_to_vector FR32:$src)),
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss)>;
+// Implicitly promote a 64-bit scalar to a vector.
+def : Pat<(v2f64 (scalar_to_vector FR64:$src)),
+ (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd)>;
-// Loading from memory automatically zeroing upper bits.
-let canFoldAsLoad = 1, isReMaterializable = 1 in
-def MOVSSrm : SSI<0x10, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
- "movss\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (loadf32 addr:$src))]>;
-
+let AddedComplexity = 20 in {
// MOVSSrm zeros the high parts of the register; represent this
// with SUBREG_TO_REG.
-let AddedComplexity = 20 in {
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector (loadf32 addr:$src))))),
(SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
def : Pat<(v4f32 (scalar_to_vector (loadf32 addr:$src))),
(SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
def : Pat<(v4f32 (X86vzmovl (loadv4f32 addr:$src))),
(SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
+// MOVSDrm zeros the high parts of the register; represent this
+// with SUBREG_TO_REG.
+def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector (loadf64 addr:$src))))),
+ (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+def : Pat<(v2f64 (scalar_to_vector (loadf64 addr:$src))),
+ (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+def : Pat<(v2f64 (X86vzmovl (loadv2f64 addr:$src))),
+ (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+def : Pat<(v2f64 (X86vzmovl (bc_v2f64 (loadv4f32 addr:$src)))),
+ (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+def : Pat<(v2f64 (X86vzload addr:$src)),
+ (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
}
// Store scalar value to memory.
def MOVSSmr : SSI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src),
"movss\t{$src, $dst|$dst, $src}",
[(store FR32:$src, addr:$dst)]>;
+def MOVSDmr : SDI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src),
+ "movsd\t{$src, $dst|$dst, $src}",
+ [(store FR64:$src, addr:$dst)]>;
// Extract and store.
def : Pat<(store (f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
addr:$dst),
(MOVSSmr addr:$dst,
(EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
+def : Pat<(store (f64 (vector_extract (v2f64 VR128:$src), (iPTR 0))),
+ addr:$dst),
+ (MOVSDmr addr:$dst,
+ (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd))>;
+
+//===----------------------------------------------------------------------===//
+// SSE 1 & 2 - Conversion Instructions
+//===----------------------------------------------------------------------===//
// Conversion instructions
def CVTTSS2SIrr : SSI<0x2C, MRMSrcReg, (outs GR32:$dst), (ins FR32:$src),
def CVTTSS2SIrm : SSI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f32mem:$src),
"cvttss2si\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (fp_to_sint (loadf32 addr:$src)))]>;
+def CVTTSD2SIrr : SDI<0x2C, MRMSrcReg, (outs GR32:$dst), (ins FR64:$src),
+ "cvttsd2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (fp_to_sint FR64:$src))]>;
+def CVTTSD2SIrm : SDI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f64mem:$src),
+ "cvttsd2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (fp_to_sint (loadf64 addr:$src)))]>;
+
def CVTSI2SSrr : SSI<0x2A, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
"cvtsi2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (sint_to_fp GR32:$src))]>;
def CVTSI2SSrm : SSI<0x2A, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
"cvtsi2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (sint_to_fp (loadi32 addr:$src)))]>;
+def CVTSI2SDrr : SDI<0x2A, MRMSrcReg, (outs FR64:$dst), (ins GR32:$src),
+ "cvtsi2sd\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (sint_to_fp GR32:$src))]>;
+def CVTSI2SDrm : SDI<0x2A, MRMSrcMem, (outs FR64:$dst), (ins i32mem:$src),
+ "cvtsi2sd\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (sint_to_fp (loadi32 addr:$src)))]>;
// Match intrinsics which expect XMM operand(s).
-def CVTSS2SIrr: SSI<0x2D, MRMSrcReg, (outs GR32:$dst), (ins FR32:$src),
- "cvtss2si{l}\t{$src, $dst|$dst, $src}", []>;
-def CVTSS2SIrm: SSI<0x2D, MRMSrcMem, (outs GR32:$dst), (ins f32mem:$src),
- "cvtss2si{l}\t{$src, $dst|$dst, $src}", []>;
-
def Int_CVTSS2SIrr : SSI<0x2D, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
"cvtss2si\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (int_x86_sse_cvtss2si VR128:$src))]>;
"cvtss2si\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (int_x86_sse_cvtss2si
(load addr:$src)))]>;
+def Int_CVTSD2SIrr : SDI<0x2D, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
+ "cvtsd2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (int_x86_sse2_cvtsd2si VR128:$src))]>;
+def Int_CVTSD2SIrm : SDI<0x2D, MRMSrcMem, (outs GR32:$dst), (ins f128mem:$src),
+ "cvtsd2si\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (int_x86_sse2_cvtsd2si
+ (load addr:$src)))]>;
// Match intrinsics which expect MM and XMM operand(s).
def Int_CVTPS2PIrr : PSI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvtps2pi\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (int_x86_sse_cvtps2pi
(load addr:$src)))]>;
+def Int_CVTPD2PIrr : PDI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
+ "cvtpd2pi\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (int_x86_sse_cvtpd2pi VR128:$src))]>;
+def Int_CVTPD2PIrm : PDI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
+ "cvtpd2pi\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (int_x86_sse_cvtpd2pi
+ (memop addr:$src)))]>;
+
+// Match intrinsics which expect MM and XMM operand(s).
def Int_CVTTPS2PIrr: PSI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvttps2pi\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (int_x86_sse_cvttps2pi VR128:$src))]>;
"cvttps2pi\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (int_x86_sse_cvttps2pi
(load addr:$src)))]>;
+def Int_CVTTPD2PIrr: PDI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
+ "cvttpd2pi\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (int_x86_sse_cvttpd2pi VR128:$src))]>;
+def Int_CVTTPD2PIrm: PDI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
+ "cvttpd2pi\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (int_x86_sse_cvttpd2pi
+ (memop addr:$src)))]>;
+
let Constraints = "$src1 = $dst" in {
def Int_CVTPI2PSrr : PSI<0x2A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR64:$src2),
(load addr:$src2)))]>;
}
-// Aliases for intrinsics
-def Int_CVTTSS2SIrr : SSI<0x2C, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
- "cvttss2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst,
- (int_x86_sse_cvttss2si VR128:$src))]>;
-def Int_CVTTSS2SIrm : SSI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f32mem:$src),
- "cvttss2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst,
- (int_x86_sse_cvttss2si(load addr:$src)))]>;
-
-let Constraints = "$src1 = $dst" in {
- def Int_CVTSI2SSrr : SSI<0x2A, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, GR32:$src2),
- "cvtsi2ss\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst, (int_x86_sse_cvtsi2ss VR128:$src1,
- GR32:$src2))]>;
- def Int_CVTSI2SSrm : SSI<0x2A, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, i32mem:$src2),
- "cvtsi2ss\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst, (int_x86_sse_cvtsi2ss VR128:$src1,
- (loadi32 addr:$src2)))]>;
-}
+//===----------------------------------------------------------------------===//
+// SSE 1 & 2 - Compare Instructions
+//===----------------------------------------------------------------------===//
// Comparison instructions
let Constraints = "$src1 = $dst", neverHasSideEffects = 1 in {
def CMPSSrr : SSIi8<0xC2, MRMSrcReg,
(outs FR32:$dst), (ins FR32:$src1, FR32:$src, SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}", []>;
-let mayLoad = 1 in
+ let mayLoad = 1 in
def CMPSSrm : SSIi8<0xC2, MRMSrcMem,
(outs FR32:$dst), (ins FR32:$src1, f32mem:$src, SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}", []>;
- // Accept explicit immediate argument form instead of comparison code.
+ def CMPSDrr : SDIi8<0xC2, MRMSrcReg,
+ (outs FR64:$dst), (ins FR64:$src1, FR64:$src, SSECC:$cc),
+ "cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
+ let mayLoad = 1 in
+ def CMPSDrm : SDIi8<0xC2, MRMSrcMem,
+ (outs FR64:$dst), (ins FR64:$src1, f64mem:$src, SSECC:$cc),
+ "cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
+
+// Accept explicit immediate argument form instead of comparison code.
let isAsmParserOnly = 1 in {
def CMPSSrr_alt : SSIi8<0xC2, MRMSrcReg,
(outs FR32:$dst), (ins FR32:$src1, FR32:$src, i8imm:$src2),
"cmpss\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
-let mayLoad = 1 in
+ let mayLoad = 1 in
def CMPSSrm_alt : SSIi8<0xC2, MRMSrcMem,
(outs FR32:$dst), (ins FR32:$src1, f32mem:$src, i8imm:$src2),
"cmpss\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
+
+ def CMPSDrr_alt : SDIi8<0xC2, MRMSrcReg,
+ (outs FR64:$dst), (ins FR64:$src1, FR64:$src, i8imm:$src2),
+ "cmpsd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
+ let mayLoad = 1 in
+ def CMPSDrm_alt : SDIi8<0xC2, MRMSrcMem,
+ (outs FR64:$dst), (ins FR64:$src1, f64mem:$src, i8imm:$src2),
+ "cmpsd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
}
}
def UCOMISSrm: PSI<0x2E, MRMSrcMem, (outs), (ins FR32:$src1, f32mem:$src2),
"ucomiss\t{$src2, $src1|$src1, $src2}",
[(set EFLAGS, (X86cmp FR32:$src1, (loadf32 addr:$src2)))]>;
-
-def COMISSrr: PSI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
- "comiss\t{$src2, $src1|$src1, $src2}", []>;
-def COMISSrm: PSI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
- "comiss\t{$src2, $src1|$src1, $src2}", []>;
-
+def UCOMISDrr: PDI<0x2E, MRMSrcReg, (outs), (ins FR64:$src1, FR64:$src2),
+ "ucomisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86cmp FR64:$src1, FR64:$src2))]>;
+def UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs), (ins FR64:$src1, f64mem:$src2),
+ "ucomisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86cmp FR64:$src1, (loadf64 addr:$src2)))]>;
} // Defs = [EFLAGS]
// Aliases to match intrinsics which expect XMM operand(s).
"cmp${cc}ss\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ss VR128:$src1,
(load addr:$src), imm:$cc))]>;
+
+ def Int_CMPSDrr : SDIi8<0xC2, MRMSrcReg,
+ (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src, SSECC:$cc),
+ "cmp${cc}sd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
+ VR128:$src, imm:$cc))]>;
+ def Int_CMPSDrm : SDIi8<0xC2, MRMSrcMem,
+ (outs VR128:$dst),
+ (ins VR128:$src1, f64mem:$src, SSECC:$cc),
+ "cmp${cc}sd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
+ (load addr:$src), imm:$cc))]>;
}
let Defs = [EFLAGS] in {
"ucomiss\t{$src2, $src1|$src1, $src2}",
[(set EFLAGS, (X86ucomi (v4f32 VR128:$src1),
(load addr:$src2)))]>;
+def Int_UCOMISDrr: PDI<0x2E, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
+ "ucomisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86ucomi (v2f64 VR128:$src1),
+ VR128:$src2))]>;
+def Int_UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs),(ins VR128:$src1, f128mem:$src2),
+ "ucomisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86ucomi (v2f64 VR128:$src1),
+ (load addr:$src2)))]>;
def Int_COMISSrr: PSI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
"comiss\t{$src2, $src1|$src1, $src2}",
"comiss\t{$src2, $src1|$src1, $src2}",
[(set EFLAGS, (X86comi (v4f32 VR128:$src1),
(load addr:$src2)))]>;
+def Int_COMISDrr: PDI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
+ "comisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86comi (v2f64 VR128:$src1),
+ VR128:$src2))]>;
+def Int_COMISDrm: PDI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
+ "comisd\t{$src2, $src1|$src1, $src2}",
+ [(set EFLAGS, (X86comi (v2f64 VR128:$src1),
+ (load addr:$src2)))]>;
} // Defs = [EFLAGS]
-// Aliases of packed SSE1 instructions for scalar use. These all have names
-// that start with 'Fs'.
+// Aliases of packed SSE1 & SSE2 instructions for scalar use. These all have
+// names that start with 'Fs'.
// Alias instructions that map fld0 to pxor for sse.
let isReMaterializable = 1, isAsCheapAsAMove = 1, isCodeGenOnly = 1,
- canFoldAsLoad = 1 in
+ canFoldAsLoad = 1 in {
// FIXME: Set encoding to pseudo!
def FsFLD0SS : I<0xEF, MRMInitReg, (outs FR32:$dst), (ins), "",
[(set FR32:$dst, fp32imm0)]>,
Requires<[HasSSE1]>, TB, OpSize;
+def FsFLD0SD : I<0xEF, MRMInitReg, (outs FR64:$dst), (ins), "",
+ [(set FR64:$dst, fpimm0)]>,
+ Requires<[HasSSE2]>, TB, OpSize;
+}
-// Alias instruction to do FR32 reg-to-reg copy using movaps. Upper bits are
-// disregarded.
-let neverHasSideEffects = 1 in
+// Alias instruction to do FR32 or FR64 reg-to-reg copy using movaps. Upper
+// bits are disregarded.
+let neverHasSideEffects = 1 in {
def FsMOVAPSrr : PSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
"movaps\t{$src, $dst|$dst, $src}", []>;
+def FsMOVAPDrr : PDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
+ "movapd\t{$src, $dst|$dst, $src}", []>;
+}
-// Alias instruction to load FR32 from f128mem using movaps. Upper bits are
-// disregarded.
-let canFoldAsLoad = 1, isReMaterializable = 1 in
+// Alias instruction to load FR32 or FR64 from f128mem using movaps. Upper
+// bits are disregarded.
+let canFoldAsLoad = 1, isReMaterializable = 1 in {
def FsMOVAPSrm : PSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src),
"movaps\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (alignedloadfsf32 addr:$src))]>;
+def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
+ "movapd\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>;
+}
/// sse12_fp_alias_pack_logical - SSE 1 & 2 aliased packed FP logical ops
///
multiclass sse12_fp_alias_pack_logical<bits<8> opc, string OpcodeStr,
- SDNode OpNode, int NoPat = 0,
- bit MayLoad = 0, bit Commutable = 1> {
- def PSrr : PSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src1, FR32:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- !if(NoPat, []<dag>,
- [(set FR32:$dst, (OpNode FR32:$src1, FR32:$src2))])> {
- let isCommutable = Commutable;
+ SDNode OpNode, bit MayLoad = 0> {
+ let isAsmParserOnly = 1 in {
+ defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode, FR32,
+ f32, f128mem, memopfsf32, SSEPackedSingle, MayLoad>, VEX_4V;
+
+ defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode, FR64,
+ f64, f128mem, memopfsf64, SSEPackedDouble, MayLoad>, OpSize,
+ VEX_4V;
}
- def PDrr : PDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- !if(NoPat, []<dag>,
- [(set FR64:$dst, (OpNode FR64:$src1, FR64:$src2))])> {
- let isCommutable = Commutable;
- }
-
- def PSrm : PSI<opc, MRMSrcMem, (outs FR32:$dst),
- (ins FR32:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- !if(NoPat, []<dag>,
- [(set FR32:$dst, (OpNode FR32:$src1,
- (memopfsf32 addr:$src2)))])> {
- let mayLoad = MayLoad;
- }
+ let Constraints = "$src1 = $dst" in {
+ defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $dst|$dst, $src2}"), OpNode, FR32, f32,
+ f128mem, memopfsf32, SSEPackedSingle, MayLoad>, TB;
- def PDrm : PDI<opc, MRMSrcMem, (outs FR64:$dst),
- (ins FR64:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- !if(NoPat, []<dag>,
- [(set FR64:$dst, (OpNode FR64:$src1,
- (memopfsf64 addr:$src2)))])> {
- let mayLoad = MayLoad;
+ defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $dst|$dst, $src2}"), OpNode, FR64, f64,
+ f128mem, memopfsf64, SSEPackedDouble, MayLoad>, TB, OpSize;
}
}
// Alias bitwise logical operations using SSE logical ops on packed FP values.
-let Constraints = "$src1 = $dst" in {
- defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand>;
- defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for>;
- defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor>;
+defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand>;
+defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for>;
+defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor>;
- let neverHasSideEffects = 1 in
- defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef, 1, 1, 0>;
-}
-
-/// sse12_fp_scalar - SSE 1 & 2 scalar instructions class
-multiclass sse12_fp_scalar<bits<8> opc, string OpcodeStr, SDNode OpNode,
- RegisterClass RC, X86MemOperand memop> {
- let isCommutable = 1 in {
- def rr : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
- OpcodeStr, [(set RC:$dst, (OpNode RC:$src1, RC:$src2))]>;
- }
- def rm : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memop:$src2),
- OpcodeStr, [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))]>;
-}
+let neverHasSideEffects = 1, Pattern = []<dag>, isCommutable = 0 in
+ defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef, 1>;
/// basic_sse12_fp_binop_rm - SSE 1 & 2 binops come in both scalar and
/// vector forms.
/// These three forms can each be reg+reg or reg+mem, so there are a total of
/// six "instructions".
///
-let Constraints = "$src1 = $dst" in {
multiclass basic_sse12_fp_binop_rm<bits<8> opc, string OpcodeStr,
- SDNode OpNode, bit Commutable = 0> {
+ SDNode OpNode> {
- let Constraints = "", isAsmParserOnly = 1, hasVEX_4VPrefix = 1 in {
- // Scalar operation, reg+reg.
- let Prefix = 12 /* XS */ in
- defm V#NAME#SS : sse12_fp_scalar<opc,
+ let isAsmParserOnly = 1 in {
+ defm V#NAME#SS : sse12_fp_scalar<opc,
!strconcat(OpcodeStr, "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- OpNode, FR32, f32mem>;
+ OpNode, FR32, f32mem>, XS, VEX_4V;
- let Prefix = 11 /* XD */ in
- defm V#NAME#SD : sse12_fp_scalar<opc,
+ defm V#NAME#SD : sse12_fp_scalar<opc,
!strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- OpNode, FR64, f64mem>;
- }
+ OpNode, FR64, f64mem>, XD, VEX_4V;
- let Constraints = "$src1 = $dst" in {
- // Scalar operation, reg+reg.
- let Prefix = 12 /* XS */ in
- defm SS : sse12_fp_scalar<opc,
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- OpNode, FR32, f32mem>;
- let Prefix = 11 /* XD */ in
- defm SD : sse12_fp_scalar<opc,
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- OpNode, FR64, f64mem>;
- }
+ defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode,
+ VR128, v4f32, f128mem, memopv4f32, SSEPackedSingle>,
+ VEX_4V;
- // Vector operation, reg+reg.
- def PSrr : PSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (v4f32 (OpNode VR128:$src1, VR128:$src2)))]> {
- let isCommutable = Commutable;
- }
+ defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode,
+ VR128, v2f64, f128mem, memopv2f64, SSEPackedDouble>,
+ OpSize, VEX_4V;
- def PDrr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (v2f64 (OpNode VR128:$src1, VR128:$src2)))]> {
- let isCommutable = Commutable;
- }
+ defm V#NAME#SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "", "_ss", ssmem, sse_load_f32>, XS, VEX_4V;
- def V#NAME#PSrr : VPSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr,
- "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []> {
- let isCommutable = Commutable;
- let Constraints = "";
- let isAsmParserOnly = 1;
+ defm V#NAME#SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "2", "_sd", sdmem, sse_load_f64>, XD, VEX_4V;
}
- def V#NAME#PDrr : VPDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr,
- "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []> {
- let isCommutable = Commutable;
- let Constraints = "";
- let isAsmParserOnly = 1;
- }
+ let Constraints = "$src1 = $dst" in {
+ defm SS : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
+ OpNode, FR32, f32mem>, XS;
- // Vector operation, reg+mem.
- def PSrm : PSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (OpNode VR128:$src1, (memopv4f32 addr:$src2)))]>;
-
- def PDrm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (OpNode VR128:$src1, (memopv2f64 addr:$src2)))]>;
-
- def V#NAME#PSrm : VPSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr,
- "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []> {
- let Constraints = "";
- let isAsmParserOnly = 1;
- }
+ defm SD : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
+ OpNode, FR64, f64mem>, XD;
- def V#NAME#PDrm : VPDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr,
- "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []> {
- let Constraints = "";
- let isAsmParserOnly = 1;
- }
+ defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $dst|$dst, $src2}"), OpNode, VR128, v4f32,
+ f128mem, memopv4f32, SSEPackedSingle>, TB;
- // Intrinsic operation, reg+reg.
- def V#NAME#SSrr_Int : VSSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse_xxx_ss
- let Constraints = "";
- }
+ defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $dst|$dst, $src2}"), OpNode, VR128, v2f64,
+ f128mem, memopv2f64, SSEPackedDouble>, TB, OpSize;
- def V#NAME#SDrr_Int : VSDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse2_xxx_sd
- let Constraints = "";
- }
-
- def SSrr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
+ defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- VR128:$src2))]>;
- // int_x86_sse_xxx_ss
+ "", "_ss", ssmem, sse_load_f32>, XS;
- def SDrr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
+ defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- VR128:$src2))]>;
- // int_x86_sse2_xxx_sd
-
- // Intrinsic operation, reg+mem.
- def V#NAME#SSrm_Int : VSSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, ssmem:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- sse_load_f32:$src2))]> {
- // int_x86_sse_xxx_ss
- let Constraints = "";
+ "2", "_sd", sdmem, sse_load_f64>, XD;
}
-
- def V#NAME#SDrm_Int : VSDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, sdmem:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- sse_load_f64:$src2))]> {
- // int_x86_sse2_xxx_sd
- let Constraints = "";
- }
-
- def SSrm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, ssmem:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- sse_load_f32:$src2))]>;
- // int_x86_sse_xxx_ss
-
- def SDrm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, sdmem:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- sse_load_f64:$src2))]>;
- // int_x86_sse2_xxx_sd
-}
}
// Arithmetic instructions
-defm ADD : basic_sse12_fp_binop_rm<0x58, "add", fadd, 1>;
-defm MUL : basic_sse12_fp_binop_rm<0x59, "mul", fmul, 1>;
+defm ADD : basic_sse12_fp_binop_rm<0x58, "add", fadd>;
+defm MUL : basic_sse12_fp_binop_rm<0x59, "mul", fmul>;
let isCommutable = 0 in {
defm SUB : basic_sse12_fp_binop_rm<0x5C, "sub", fsub>;
///
/// This provides a total of eight "instructions".
///
-let Constraints = "$src1 = $dst" in {
multiclass sse12_fp_binop_rm<bits<8> opc, string OpcodeStr,
- SDNode OpNode, bit Commutable = 0> {
-
- // Scalar operation, reg+reg.
- def SSrr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src1, FR32:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set FR32:$dst, (OpNode FR32:$src1, FR32:$src2))]> {
- let isCommutable = Commutable;
- }
+ SDNode OpNode> {
- def SDrr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set FR64:$dst, (OpNode FR64:$src1, FR64:$src2))]> {
- let isCommutable = Commutable;
- }
-
- // Scalar operation, reg+mem.
- def SSrm : SSI<opc, MRMSrcMem, (outs FR32:$dst),
- (ins FR32:$src1, f32mem:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set FR32:$dst, (OpNode FR32:$src1, (load addr:$src2)))]>;
-
- def SDrm : SDI<opc, MRMSrcMem, (outs FR64:$dst),
- (ins FR64:$src1, f64mem:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set FR64:$dst, (OpNode FR64:$src1, (load addr:$src2)))]>;
-
- // Vector operation, reg+reg.
- def PSrr : PSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (v4f32 (OpNode VR128:$src1, VR128:$src2)))]> {
- let isCommutable = Commutable;
- }
-
- def PDrr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (v2f64 (OpNode VR128:$src1, VR128:$src2)))]> {
- let isCommutable = Commutable;
- }
-
- // Vector operation, reg+mem.
- def PSrm : PSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (OpNode VR128:$src1, (memopv4f32 addr:$src2)))]>;
-
- def PDrm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (OpNode VR128:$src1, (memopv2f64 addr:$src2)))]>;
-
- // Intrinsic operation, reg+reg.
- def SSrr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse_xxx_ss
- let isCommutable = Commutable;
- }
-
- def SDrr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse2_xxx_sd
- let isCommutable = Commutable;
+ let isAsmParserOnly = 1 in {
+ // Scalar operation, reg+reg.
+ defm V#NAME#SS : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ OpNode, FR32, f32mem>, XS, VEX_4V;
+
+ defm V#NAME#SD : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ OpNode, FR64, f64mem>, XD, VEX_4V;
+
+ defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode,
+ VR128, v4f32, f128mem, memopv4f32, SSEPackedSingle>,
+ VEX_4V;
+
+ defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), OpNode,
+ VR128, v2f64, f128mem, memopv2f64, SSEPackedDouble>,
+ OpSize, VEX_4V;
+
+ defm V#NAME#SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "", "_ss", ssmem, sse_load_f32>, XS, VEX_4V;
+
+ defm V#NAME#SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "2", "_sd", sdmem, sse_load_f64>, XD, VEX_4V;
+
+ defm V#NAME#PS : sse12_fp_packed_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "", "_ps", f128mem, memopv4f32, SSEPackedSingle>, VEX_4V;
+
+ defm V#NAME#PD : sse12_fp_packed_int<opc, OpcodeStr, VR128,
+ !strconcat(OpcodeStr, "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "2", "_pd", f128mem, memopv2f64, SSEPackedDouble>, OpSize,
+ VEX_4V;
}
- // Intrinsic operation, reg+mem.
- def SSrm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, ssmem:$src2),
+ let Constraints = "$src1 = $dst" in {
+ // Scalar operation, reg+reg.
+ defm SS : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
+ OpNode, FR32, f32mem>, XS;
+ defm SD : sse12_fp_scalar<opc,
+ !strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
+ OpNode, FR64, f64mem>, XD;
+ defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "ps\t{$src2, $dst|$dst, $src2}"), OpNode, VR128, v4f32,
+ f128mem, memopv4f32, SSEPackedSingle>, TB;
+
+ defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr,
+ "pd\t{$src2, $dst|$dst, $src2}"), OpNode, VR128, v2f64,
+ f128mem, memopv2f64, SSEPackedDouble>, TB, OpSize;
+
+ defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ss")) VR128:$src1,
- sse_load_f32:$src2))]>;
- // int_x86_sse_xxx_ss
+ "", "_ss", ssmem, sse_load_f32>, XS;
- def SDrm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, sdmem:$src2),
+ defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_sd")) VR128:$src1,
- sse_load_f64:$src2))]>;
- // int_x86_sse2_xxx_sd
+ "2", "_sd", sdmem, sse_load_f64>, XD;
- // Vector intrinsic operation, reg+reg.
- def PSrr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
+ defm PS : sse12_fp_packed_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ps")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse_xxx_ps
- let isCommutable = Commutable;
- }
+ "", "_ps", f128mem, memopv4f32, SSEPackedSingle>, TB;
- def PDrr_Int : PDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
+ defm PD : sse12_fp_packed_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_pd")) VR128:$src1,
- VR128:$src2))]> {
- // int_x86_sse2_xxx_pd
- let isCommutable = Commutable;
+ "2", "_pd", f128mem, memopv2f64, SSEPackedDouble>, TB, OpSize;
}
-
- // Vector intrinsic operation, reg+mem.
- def PSrm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse_",
- !strconcat(OpcodeStr, "_ps")) VR128:$src1,
- (memopv4f32 addr:$src2)))]>;
- // int_x86_sse_xxx_ps
-
- def PDrm_Int : PDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (!nameconcat<Intrinsic>("int_x86_sse2_",
- !strconcat(OpcodeStr, "_pd")) VR128:$src1,
- (memopv2f64 addr:$src2)))]>;
- // int_x86_sse2_xxx_pd
-}
}
-defm MAX : sse12_fp_binop_rm<0x5F, "max", X86fmax>;
-defm MIN : sse12_fp_binop_rm<0x5D, "min", X86fmin>;
+let isCommutable = 0 in {
+ defm MAX : sse12_fp_binop_rm<0x5F, "max", X86fmax>;
+ defm MIN : sse12_fp_binop_rm<0x5D, "min", X86fmin>;
+}
//===----------------------------------------------------------------------===//
// SSE packed FP Instructions
defm RCP : sse1_fp_unop_rm<0x53, "rcp", X86frcp,
int_x86_sse_rcp_ss, int_x86_sse_rcp_ps>;
-/// sse12_fp_pack_logical - SSE 1 & 2 packed FP logical ops
+/// sse12_fp_packed_logical - SSE 1 & 2 packed FP logical ops
///
-multiclass sse12_fp_pack_logical<bits<8> opc, string OpcodeStr,
+multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
SDNode OpNode, int HasPat = 0,
- bit Commutable = 1,
list<list<dag>> Pattern = []> {
- def PSrr : PSI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- !if(HasPat, Pattern[0],
- [(set VR128:$dst, (v2i64 (OpNode VR128:$src1,
- VR128:$src2)))])>
- { let isCommutable = Commutable; }
-
- def PDrr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- !if(HasPat, Pattern[1],
- [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
- (bc_v2i64 (v2f64 VR128:$src2))))])>
- { let isCommutable = Commutable; }
-
- def PSrm : PSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"),
- !if(HasPat, Pattern[2],
- [(set VR128:$dst, (OpNode (bc_v2i64 (v4f32 VR128:$src1)),
- (memopv2i64 addr:$src2)))])>;
-
- def PDrm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, f128mem:$src2),
- !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"),
- !if(HasPat, Pattern[3],
- [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
- (memopv2i64 addr:$src2)))])>;
+ let isAsmParserOnly = 1 in {
+ defm V#NAME#PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
+ !strconcat(OpcodeStr, "ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ f128mem,
+ !if(HasPat, Pattern[0], // rr
+ [(set VR128:$dst, (v2i64 (OpNode VR128:$src1,
+ VR128:$src2)))]),
+ !if(HasPat, Pattern[2], // rm
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v4f32 VR128:$src1)),
+ (memopv2i64 addr:$src2)))])>,
+ VEX_4V;
+
+ defm V#NAME#PD : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedDouble,
+ !strconcat(OpcodeStr, "pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ f128mem,
+ !if(HasPat, Pattern[1], // rr
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
+ (bc_v2i64 (v2f64
+ VR128:$src2))))]),
+ !if(HasPat, Pattern[3], // rm
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
+ (memopv2i64 addr:$src2)))])>,
+ OpSize, VEX_4V;
+ }
+ let Constraints = "$src1 = $dst" in {
+ defm PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
+ !strconcat(OpcodeStr, "ps\t{$src2, $dst|$dst, $src2}"), f128mem,
+ !if(HasPat, Pattern[0], // rr
+ [(set VR128:$dst, (v2i64 (OpNode VR128:$src1,
+ VR128:$src2)))]),
+ !if(HasPat, Pattern[2], // rm
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v4f32 VR128:$src1)),
+ (memopv2i64 addr:$src2)))])>, TB;
+
+ defm PD : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedDouble,
+ !strconcat(OpcodeStr, "pd\t{$src2, $dst|$dst, $src2}"), f128mem,
+ !if(HasPat, Pattern[1], // rr
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
+ (bc_v2i64 (v2f64
+ VR128:$src2))))]),
+ !if(HasPat, Pattern[3], // rm
+ [(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
+ (memopv2i64 addr:$src2)))])>,
+ TB, OpSize;
+ }
}
// Logical
-let Constraints = "$src1 = $dst" in {
- defm AND : sse12_fp_pack_logical<0x54, "and", and>;
- defm OR : sse12_fp_pack_logical<0x56, "or", or>;
- defm XOR : sse12_fp_pack_logical<0x57, "xor", xor>;
- defm ANDN : sse12_fp_pack_logical<0x55, "andn", undef /* dummy */, 1, 0, [
+defm AND : sse12_fp_packed_logical<0x54, "and", and>;
+defm OR : sse12_fp_packed_logical<0x56, "or", or>;
+defm XOR : sse12_fp_packed_logical<0x57, "xor", xor>;
+let isCommutable = 0 in
+ defm ANDN : sse12_fp_packed_logical<0x55, "andn", undef /* dummy */, 1, [
// single r+r
[(set VR128:$dst, (v2i64 (and (xor VR128:$src1,
(bc_v2i64 (v4i32 immAllOnesV))),
// double r+m
[(set VR128:$dst, (and (vnot (bc_v2i64 (v2f64 VR128:$src1))),
(memopv2i64 addr:$src2)))]]>;
-}
let Constraints = "$src1 = $dst" in {
def CMPPSrri : PSIi8<0xC2, MRMSrcReg,
"cmp${cc}ps\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ps VR128:$src1,
(memop addr:$src), imm:$cc))]>;
+ def CMPPDrri : PDIi8<0xC2, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
+ "cmp${cc}pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
+ VR128:$src, imm:$cc))]>;
+ def CMPPDrmi : PDIi8<0xC2, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f128mem:$src, SSECC:$cc),
+ "cmp${cc}pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
+ (memop addr:$src), imm:$cc))]>;
// Accept explicit immediate argument form instead of comparison code.
let isAsmParserOnly = 1 in {
def CMPPSrmi_alt : PSIi8<0xC2, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src, i8imm:$src2),
"cmpps\t{$src2, $src, $dst|$dst, $src, $src}", []>;
+ def CMPPDrri_alt : PDIi8<0xC2, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src, i8imm:$src2),
+ "cmppd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
+ def CMPPDrmi_alt : PDIi8<0xC2, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f128mem:$src, i8imm:$src2),
+ "cmppd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
}
}
def : Pat<(v4i32 (X86cmpps (v4f32 VR128:$src1), VR128:$src2, imm:$cc)),
(CMPPSrri (v4f32 VR128:$src1), (v4f32 VR128:$src2), imm:$cc)>;
def : Pat<(v4i32 (X86cmpps (v4f32 VR128:$src1), (memop addr:$src2), imm:$cc)),
(CMPPSrmi (v4f32 VR128:$src1), addr:$src2, imm:$cc)>;
+def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), VR128:$src2, imm:$cc)),
+ (CMPPDrri VR128:$src1, VR128:$src2, imm:$cc)>;
+def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
+ (CMPPDrmi VR128:$src1, addr:$src2, imm:$cc)>;
// Shuffle and unpack instructions
let Constraints = "$src1 = $dst" in {
[(set VR128:$dst,
(v4f32 (shufp:$src3
VR128:$src1, (memopv4f32 addr:$src2))))]>;
+ def SHUFPDrri : PDIi8<0xC6, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+ [(set VR128:$dst,
+ (v2f64 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
+ def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1,
+ f128mem:$src2, i8imm:$src3),
+ "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+ [(set VR128:$dst,
+ (v2f64 (shufp:$src3
+ VR128:$src1, (memopv2f64 addr:$src2))))]>;
let AddedComplexity = 10 in {
def UNPCKHPSrr : PSI<0x15, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
"unpcklps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
- (unpckl VR128:$src1, (memopv4f32 addr:$src2)))]>;
+ (unpckl VR128:$src1, (memopv4f32 addr:$src2)))]>;
+ def UNPCKHPDrr : PDI<0x15, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ "unpckhpd\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst,
+ (v2f64 (unpckh VR128:$src1, VR128:$src2)))]>;
+ def UNPCKHPDrm : PDI<0x15, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
+ "unpckhpd\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst,
+ (v2f64 (unpckh VR128:$src1,
+ (memopv2f64 addr:$src2))))]>;
+
+ def UNPCKLPDrr : PDI<0x14, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ "unpcklpd\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst,
+ (v2f64 (unpckl VR128:$src1, VR128:$src2)))]>;
+ def UNPCKLPDrm : PDI<0x14, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
+ "unpcklpd\t{$src2, $dst|$dst, $src2}",
+ [(set VR128:$dst,
+ (unpckl VR128:$src1, (memopv2f64 addr:$src2)))]>;
} // AddedComplexity
} // Constraints = "$src1 = $dst"
// SSE2 Instructions
//===---------------------------------------------------------------------===//
-// Move Instructions. Register-to-register movsd is not used for FR64
-// register copies because it's a partial register update; FsMOVAPDrr is
-// used instead. Register-to-register movsd is not modeled as an INSERT_SUBREG
-// because INSERT_SUBREG requires that the insert be implementable in terms of
-// a copy, and just mentioned, we don't use movsd for copies.
-let Constraints = "$src1 = $dst" in
-def MOVSDrr : SDI<0x10, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, FR64:$src2),
- "movsd\t{$src2, $dst|$dst, $src2}",
- [(set (v2f64 VR128:$dst),
- (movl VR128:$src1, (scalar_to_vector FR64:$src2)))]>;
-
-// Extract the low 64-bit value from one vector and insert it into another.
-let AddedComplexity = 15 in
-def : Pat<(v2f64 (movl VR128:$src1, VR128:$src2)),
- (MOVSDrr (v2f64 VR128:$src1),
- (EXTRACT_SUBREG (v2f64 VR128:$src2), sub_sd))>;
-
-// Implicitly promote a 64-bit scalar to a vector.
-def : Pat<(v2f64 (scalar_to_vector FR64:$src)),
- (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd)>;
-
-// Loading from memory automatically zeroing upper bits.
-let canFoldAsLoad = 1, isReMaterializable = 1, AddedComplexity = 20 in
-def MOVSDrm : SDI<0x10, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
- "movsd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (loadf64 addr:$src))]>;
-
-// MOVSDrm zeros the high parts of the register; represent this
-// with SUBREG_TO_REG.
-let AddedComplexity = 20 in {
-def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector (loadf64 addr:$src))))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
-def : Pat<(v2f64 (scalar_to_vector (loadf64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
-def : Pat<(v2f64 (X86vzmovl (loadv2f64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
-def : Pat<(v2f64 (X86vzmovl (bc_v2f64 (loadv4f32 addr:$src)))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
-def : Pat<(v2f64 (X86vzload addr:$src)),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
-}
-
-// Store scalar value to memory.
-def MOVSDmr : SDI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src),
- "movsd\t{$src, $dst|$dst, $src}",
- [(store FR64:$src, addr:$dst)]>;
-
-// Extract and store.
-def : Pat<(store (f64 (vector_extract (v2f64 VR128:$src), (iPTR 0))),
- addr:$dst),
- (MOVSDmr addr:$dst,
- (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd))>;
-
// Conversion instructions
-def CVTTSD2SIrr : SDI<0x2C, MRMSrcReg, (outs GR32:$dst), (ins FR64:$src),
- "cvttsd2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (fp_to_sint FR64:$src))]>;
-def CVTTSD2SIrm : SDI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f64mem:$src),
- "cvttsd2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (fp_to_sint (loadf64 addr:$src)))]>;
def CVTSD2SSrr : SDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src),
"cvtsd2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (fround FR64:$src))]>;
"cvtsd2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (fround (loadf64 addr:$src)))]>, XD,
Requires<[HasSSE2, OptForSize]>;
-def CVTSI2SDrr : SDI<0x2A, MRMSrcReg, (outs FR64:$dst), (ins GR32:$src),
- "cvtsi2sd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (sint_to_fp GR32:$src))]>;
-def CVTSI2SDrm : SDI<0x2A, MRMSrcMem, (outs FR64:$dst), (ins i32mem:$src),
- "cvtsi2sd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (sint_to_fp (loadi32 addr:$src)))]>;
-def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
-def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
-def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
-def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}", []>;
def CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}", []>;
-def CVTDQ2PSrr : PSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
-def CVTDQ2PSrm : PSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
def COMISDrr: PDI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
"comisd\t{$src2, $src1|$src1, $src2}", []>;
def COMISDrm: PDI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
(CVTSS2SDrr (MOVSSrm addr:$src))>,
Requires<[HasSSE2, OptForSpeed]>;
-// Match intrinsics which expect XMM operand(s).
-def Int_CVTSD2SIrr : SDI<0x2D, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
- "cvtsd2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_cvtsd2si VR128:$src))]>;
-def Int_CVTSD2SIrm : SDI<0x2D, MRMSrcMem, (outs GR32:$dst), (ins f128mem:$src),
- "cvtsd2si\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_cvtsd2si
- (load addr:$src)))]>;
-
// Match intrinsics which expect MM and XMM operand(s).
-def Int_CVTPD2PIrr : PDI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
- "cvtpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvtpd2pi VR128:$src))]>;
-def Int_CVTPD2PIrm : PDI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
- "cvtpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvtpd2pi
- (memop addr:$src)))]>;
-def Int_CVTTPD2PIrr: PDI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
- "cvttpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvttpd2pi VR128:$src))]>;
-def Int_CVTTPD2PIrm: PDI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
- "cvttpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvttpd2pi
- (memop addr:$src)))]>;
def Int_CVTPI2PDrr : PDI<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src),
"cvtpi2pd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cvtpi2pd VR64:$src))]>;
[(set GR32:$dst, (int_x86_sse2_cvttsd2si
(load addr:$src)))]>;
-// Comparison instructions
-let Constraints = "$src1 = $dst", neverHasSideEffects = 1 in {
- def CMPSDrr : SDIi8<0xC2, MRMSrcReg,
- (outs FR64:$dst), (ins FR64:$src1, FR64:$src, SSECC:$cc),
- "cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
-let mayLoad = 1 in
- def CMPSDrm : SDIi8<0xC2, MRMSrcMem,
- (outs FR64:$dst), (ins FR64:$src1, f64mem:$src, SSECC:$cc),
- "cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
-
- // Accept explicit immediate argument form instead of comparison code.
-let isAsmParserOnly = 1 in {
- def CMPSDrr_alt : SDIi8<0xC2, MRMSrcReg,
- (outs FR64:$dst), (ins FR64:$src1, FR64:$src, i8imm:$src2),
- "cmpsd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
-let mayLoad = 1 in
- def CMPSDrm_alt : SDIi8<0xC2, MRMSrcMem,
- (outs FR64:$dst), (ins FR64:$src1, f64mem:$src, i8imm:$src2),
- "cmpsd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
-}
-}
-
-let Defs = [EFLAGS] in {
-def UCOMISDrr: PDI<0x2E, MRMSrcReg, (outs), (ins FR64:$src1, FR64:$src2),
- "ucomisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86cmp FR64:$src1, FR64:$src2))]>;
-def UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs), (ins FR64:$src1, f64mem:$src2),
- "ucomisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86cmp FR64:$src1, (loadf64 addr:$src2)))]>;
-} // Defs = [EFLAGS]
-
-// Aliases to match intrinsics which expect XMM operand(s).
-let Constraints = "$src1 = $dst" in {
- def Int_CMPSDrr : SDIi8<0xC2, MRMSrcReg,
- (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src, SSECC:$cc),
- "cmp${cc}sd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
- VR128:$src, imm:$cc))]>;
- def Int_CMPSDrm : SDIi8<0xC2, MRMSrcMem,
- (outs VR128:$dst),
- (ins VR128:$src1, f64mem:$src, SSECC:$cc),
- "cmp${cc}sd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
- (load addr:$src), imm:$cc))]>;
-}
-
-let Defs = [EFLAGS] in {
-def Int_UCOMISDrr: PDI<0x2E, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
- "ucomisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86ucomi (v2f64 VR128:$src1),
- VR128:$src2))]>;
-def Int_UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs),(ins VR128:$src1, f128mem:$src2),
- "ucomisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86ucomi (v2f64 VR128:$src1),
- (load addr:$src2)))]>;
-
-def Int_COMISDrr: PDI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
- "comisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86comi (v2f64 VR128:$src1),
- VR128:$src2))]>;
-def Int_COMISDrm: PDI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
- "comisd\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86comi (v2f64 VR128:$src1),
- (load addr:$src2)))]>;
-} // Defs = [EFLAGS]
-
-// Aliases of packed SSE2 instructions for scalar use. These all have names
-// that start with 'Fs'.
-
-// Alias instructions that map fld0 to pxor for sse.
-let isReMaterializable = 1, isAsCheapAsAMove = 1, isCodeGenOnly = 1,
- canFoldAsLoad = 1 in
-def FsFLD0SD : I<0xEF, MRMInitReg, (outs FR64:$dst), (ins), "",
- [(set FR64:$dst, fpimm0)]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-// Alias instruction to do FR64 reg-to-reg copy using movapd. Upper bits are
-// disregarded.
-let neverHasSideEffects = 1 in
-def FsMOVAPDrr : PDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>;
-
-// Alias instruction to load FR64 from f128mem using movapd. Upper bits are
-// disregarded.
-let canFoldAsLoad = 1, isReMaterializable = 1 in
-def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
- "movapd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>;
-
//===---------------------------------------------------------------------===//
// SSE packed FP Instructions
// There is no f64 version of the reciprocal approximation instructions.
-let Constraints = "$src1 = $dst" in {
- def CMPPDrri : PDIi8<0xC2, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
- "cmp${cc}pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
- VR128:$src, imm:$cc))]>;
- def CMPPDrmi : PDIi8<0xC2, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f128mem:$src, SSECC:$cc),
- "cmp${cc}pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
- (memop addr:$src), imm:$cc))]>;
-
- // Accept explicit immediate argument form instead of comparison code.
-let isAsmParserOnly = 1 in {
- def CMPPDrri_alt : PDIi8<0xC2, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src, i8imm:$src2),
- "cmppd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
- def CMPPDrmi_alt : PDIi8<0xC2, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f128mem:$src, i8imm:$src2),
- "cmppd\t{$src2, $src, $dst|$dst, $src, $src2}", []>;
-}
-}
-def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), VR128:$src2, imm:$cc)),
- (CMPPDrri VR128:$src1, VR128:$src2, imm:$cc)>;
-def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
- (CMPPDrmi VR128:$src1, addr:$src2, imm:$cc)>;
-
-// Shuffle and unpack instructions
-let Constraints = "$src1 = $dst" in {
- def SHUFPDrri : PDIi8<0xC6, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src2, i8imm:$src3),
- "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- [(set VR128:$dst,
- (v2f64 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
- def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1,
- f128mem:$src2, i8imm:$src3),
- "shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- [(set VR128:$dst,
- (v2f64 (shufp:$src3
- VR128:$src1, (memopv2f64 addr:$src2))))]>;
-
- let AddedComplexity = 10 in {
- def UNPCKHPDrr : PDI<0x15, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
- "unpckhpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
- (v2f64 (unpckh VR128:$src1, VR128:$src2)))]>;
- def UNPCKHPDrm : PDI<0x15, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
- "unpckhpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
- (v2f64 (unpckh VR128:$src1,
- (memopv2f64 addr:$src2))))]>;
-
- def UNPCKLPDrr : PDI<0x14, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
- "unpcklpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
- (v2f64 (unpckl VR128:$src1, VR128:$src2)))]>;
- def UNPCKLPDrm : PDI<0x14, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
- "unpcklpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
- (unpckl VR128:$src1, (memopv2f64 addr:$src2)))]>;
- } // AddedComplexity
-} // Constraints = "$src1 = $dst"
-
-
//===---------------------------------------------------------------------===//
// SSE integer instructions
let ExeDomain = SSEPackedInt in {
// SSE3 Instructions
//===---------------------------------------------------------------------===//
+// Conversion Instructions
+def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
+
// Move Instructions
def MOVSHDUPrr : S3SI<0x16, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movshdup\t{$src, $dst|$dst, $src}",
projects / oota-llvm.git / blobdiff