diff options
| -rw-r--r-- | lib/Target/X86/X86InstrSSE.td | 2138 | ||||
| -rw-r--r-- | lib/Target/X86/X86Schedule.td | 136 | ||||
| -rw-r--r-- | lib/Target/X86/X86ScheduleAtom.td | 151 |
3 files changed, 1697 insertions, 728 deletions
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td index 56e997c..c6d1d19 100644 --- a/lib/Target/X86/X86InstrSSE.td +++ b/lib/Target/X86/X86InstrSSE.td @@ -13,6 +13,126 @@ // //===----------------------------------------------------------------------===// +class OpndItins<InstrItinClass arg_rr, InstrItinClass arg_rm> { + InstrItinClass rr = arg_rr; + InstrItinClass rm = arg_rm; +} + +class SizeItins<OpndItins arg_s, OpndItins arg_d> { + OpndItins s = arg_s; + OpndItins d = arg_d; +} + + +class ShiftOpndItins<InstrItinClass arg_rr, InstrItinClass arg_rm, + InstrItinClass arg_ri> { + InstrItinClass rr = arg_rr; + InstrItinClass rm = arg_rm; + InstrItinClass ri = arg_ri; +} + + +// scalar +def SSE_ALU_F32S : OpndItins< + IIC_SSE_ALU_F32S_RR, IIC_SSE_ALU_F32S_RM +>; + +def SSE_ALU_F64S : OpndItins< + IIC_SSE_ALU_F64S_RR, IIC_SSE_ALU_F64S_RM +>; + +def SSE_ALU_ITINS_S : SizeItins< + SSE_ALU_F32S, SSE_ALU_F64S +>; + +def SSE_MUL_F32S : OpndItins< + IIC_SSE_MUL_F32S_RR, IIC_SSE_MUL_F64S_RM +>; + +def SSE_MUL_F64S : OpndItins< + IIC_SSE_MUL_F64S_RR, IIC_SSE_MUL_F64S_RM +>; + +def SSE_MUL_ITINS_S : SizeItins< + SSE_MUL_F32S, SSE_MUL_F64S +>; + +def SSE_DIV_F32S : OpndItins< + IIC_SSE_DIV_F32S_RR, IIC_SSE_DIV_F64S_RM +>; + +def SSE_DIV_F64S : OpndItins< + IIC_SSE_DIV_F64S_RR, IIC_SSE_DIV_F64S_RM +>; + +def SSE_DIV_ITINS_S : SizeItins< + SSE_DIV_F32S, SSE_DIV_F64S +>; + +// parallel +def SSE_ALU_F32P : OpndItins< + IIC_SSE_ALU_F32P_RR, IIC_SSE_ALU_F32P_RM +>; + +def SSE_ALU_F64P : OpndItins< + IIC_SSE_ALU_F64P_RR, IIC_SSE_ALU_F64P_RM +>; + +def SSE_ALU_ITINS_P : SizeItins< + SSE_ALU_F32P, SSE_ALU_F64P +>; + +def SSE_MUL_F32P : OpndItins< + IIC_SSE_MUL_F32P_RR, IIC_SSE_MUL_F64P_RM +>; + +def SSE_MUL_F64P : OpndItins< + IIC_SSE_MUL_F64P_RR, IIC_SSE_MUL_F64P_RM +>; + +def SSE_MUL_ITINS_P : SizeItins< + SSE_MUL_F32P, SSE_MUL_F64P +>; + +def SSE_DIV_F32P : OpndItins< + IIC_SSE_DIV_F32P_RR, IIC_SSE_DIV_F64P_RM +>; + +def SSE_DIV_F64P : OpndItins< + IIC_SSE_DIV_F64P_RR, IIC_SSE_DIV_F64P_RM +>; + +def SSE_DIV_ITINS_P : SizeItins< + SSE_DIV_F32P, SSE_DIV_F64P +>; + +def SSE_BIT_ITINS_P : OpndItins< + IIC_SSE_BIT_P_RR, IIC_SSE_BIT_P_RM +>; + +def SSE_INTALU_ITINS_P : OpndItins< + IIC_SSE_INTALU_P_RR, IIC_SSE_INTALU_P_RM +>; + +def SSE_INTALUQ_ITINS_P : OpndItins< + IIC_SSE_INTALUQ_P_RR, IIC_SSE_INTALUQ_P_RM +>; + +def SSE_INTMUL_ITINS_P : OpndItins< + IIC_SSE_INTMUL_P_RR, IIC_SSE_INTMUL_P_RM +>; + +def SSE_INTSHIFT_ITINS_P : ShiftOpndItins< + IIC_SSE_INTSH_P_RR, IIC_SSE_INTSH_P_RM, IIC_SSE_INTSH_P_RI +>; + +def SSE_MOVA_ITINS : OpndItins< + IIC_SSE_MOVA_P_RR, IIC_SSE_MOVA_P_RM +>; + +def SSE_MOVU_ITINS : OpndItins< + IIC_SSE_MOVU_P_RR, IIC_SSE_MOVU_P_RM +>; //===----------------------------------------------------------------------===// // SSE 1 & 2 Instructions Classes @@ -21,25 +141,27 @@ /// sse12_fp_scalar - SSE 1 & 2 scalar instructions class multiclass sse12_fp_scalar<bits<8> opc, string OpcodeStr, SDNode OpNode, RegisterClass RC, X86MemOperand x86memop, + OpndItins itins, bit Is2Addr = 1> { let isCommutable = 1 in { def rr : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (OpNode RC:$src1, RC:$src2))]>; + [(set RC:$dst, (OpNode RC:$src1, RC:$src2))], itins.rr>; } def rm : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))]>; + [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))], itins.rm>; } /// sse12_fp_scalar_int - SSE 1 & 2 scalar instructions intrinsics class multiclass sse12_fp_scalar_int<bits<8> opc, string OpcodeStr, RegisterClass RC, string asm, string SSEVer, string FPSizeStr, Operand memopr, ComplexPattern mem_cpat, + OpndItins itins, bit Is2Addr = 1> { def rr_Int : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, @@ -47,34 +169,34 @@ multiclass sse12_fp_scalar_int<bits<8> opc, string OpcodeStr, RegisterClass RC, !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (!cast<Intrinsic>( !strconcat("int_x86_sse", SSEVer, "_", OpcodeStr, FPSizeStr)) - RC:$src1, RC:$src2))]>; + RC:$src1, RC:$src2))], itins.rr>; def rm_Int : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memopr:$src2), !if(Is2Addr, !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"), !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (!cast<Intrinsic>(!strconcat("int_x86_sse", SSEVer, "_", OpcodeStr, FPSizeStr)) - RC:$src1, mem_cpat:$src2))]>; + RC:$src1, mem_cpat:$src2))], itins.rm>; } /// 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 Is2Addr = 1> { + Domain d, OpndItins itins, bit Is2Addr = 1> { let isCommutable = 1 in def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_DEFAULT, d>; + [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], itins.rr, d>; let mayLoad = 1 in def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (OpNode RC:$src1, (mem_frag addr:$src2)))], - IIC_DEFAULT, d>; + itins.rm, d>; } /// sse12_fp_packed_logical_rm - SSE 1 & 2 packed instructions class @@ -100,7 +222,7 @@ multiclass sse12_fp_packed_logical_rm<bits<8> opc, RegisterClass RC, Domain d, 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, bit Is2Addr = 1> { + Domain d, OpndItins itins, bit Is2Addr = 1> { def rr_Int : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"), @@ -348,14 +470,16 @@ let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1, class sse12_move_rr<RegisterClass RC, SDNode OpNode, ValueType vt, string asm> : SI<0x10, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, RC:$src2), asm, [(set VR128:$dst, (vt (OpNode VR128:$src1, - (scalar_to_vector RC:$src2))))]>; + (scalar_to_vector RC:$src2))))], + IIC_SSE_MOV_S_RR>; // Loading from memory automatically zeroing upper bits. class sse12_move_rm<RegisterClass RC, X86MemOperand x86memop, PatFrag mem_pat, string OpcodeStr> : SI<0x10, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), - [(set RC:$dst, (mem_pat addr:$src))]>; + [(set RC:$dst, (mem_pat addr:$src))], + IIC_SSE_MOV_S_RM>; // AVX def VMOVSSrr : sse12_move_rr<FR32, X86Movss, v4f32, @@ -369,11 +493,13 @@ def VMOVSDrr : sse12_move_rr<FR64, X86Movsd, v2f64, let isCodeGenOnly = 1 in { def VMOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src1, FR32:$src2), - "movss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, + "movss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [], + IIC_SSE_MOV_S_RR>, XS, VEX_4V, VEX_LIG; def VMOVSDrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src1, FR64:$src2), - "movsd\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, + "movsd\t{$src2, $src1, $dst|$dst, $src1, $src2}", [], + IIC_SSE_MOV_S_RR>, XD, VEX_4V, VEX_LIG; } @@ -387,10 +513,12 @@ let canFoldAsLoad = 1, isReMaterializable = 1 in { def VMOVSSmr : SI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src), "movss\t{$src, $dst|$dst, $src}", - [(store FR32:$src, addr:$dst)]>, XS, VEX, VEX_LIG; + [(store FR32:$src, addr:$dst)], IIC_SSE_MOV_S_MR>, + XS, VEX, VEX_LIG; def VMOVSDmr : SI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src), "movsd\t{$src, $dst|$dst, $src}", - [(store FR64:$src, addr:$dst)]>, XD, VEX, VEX_LIG; + [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>, + XD, VEX, VEX_LIG; // SSE1 & 2 let Constraints = "$src1 = $dst" in { @@ -403,10 +531,12 @@ let Constraints = "$src1 = $dst" in { let isCodeGenOnly = 1 in { def MOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src1, FR32:$src2), - "movss\t{$src2, $dst|$dst, $src2}", []>, XS; + "movss\t{$src2, $dst|$dst, $src2}", [], + IIC_SSE_MOV_S_RR>, XS; def MOVSDrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src1, FR64:$src2), - "movsd\t{$src2, $dst|$dst, $src2}", []>, XD; + "movsd\t{$src2, $dst|$dst, $src2}", [], + IIC_SSE_MOV_S_RR>, XD; } } @@ -419,10 +549,10 @@ let canFoldAsLoad = 1, isReMaterializable = 1 in { def MOVSSmr : SSI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src), "movss\t{$src, $dst|$dst, $src}", - [(store FR32:$src, addr:$dst)]>; + [(store FR32:$src, addr:$dst)], IIC_SSE_MOV_S_MR>; def MOVSDmr : SDI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src), "movsd\t{$src, $dst|$dst, $src}", - [(store FR64:$src, addr:$dst)]>; + [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>; // Patterns let Predicates = [HasAVX] in { @@ -684,93 +814,122 @@ let Predicates = [HasSSE2] in { multiclass sse12_mov_packed<bits<8> opc, RegisterClass RC, X86MemOperand x86memop, PatFrag ld_frag, string asm, Domain d, + OpndItins itins, bit IsReMaterializable = 1> { let neverHasSideEffects = 1 in def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src), - !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], IIC_DEFAULT, d>; + !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], itins.rr, d>; let canFoldAsLoad = 1, isReMaterializable = IsReMaterializable in def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src), !strconcat(asm, "\t{$src, $dst|$dst, $src}"), - [(set RC:$dst, (ld_frag addr:$src))], IIC_DEFAULT, d>; + [(set RC:$dst, (ld_frag addr:$src))], itins.rm, d>; } defm VMOVAPS : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv4f32, - "movaps", SSEPackedSingle>, TB, VEX; + "movaps", SSEPackedSingle, SSE_MOVA_ITINS>, + TB, VEX; defm VMOVAPD : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv2f64, - "movapd", SSEPackedDouble>, TB, OpSize, VEX; + "movapd", SSEPackedDouble, SSE_MOVA_ITINS>, + TB, OpSize, VEX; defm VMOVUPS : sse12_mov_packed<0x10, VR128, f128mem, loadv4f32, - "movups", SSEPackedSingle>, TB, VEX; + "movups", SSEPackedSingle, SSE_MOVU_ITINS>, + TB, VEX; defm VMOVUPD : sse12_mov_packed<0x10, VR128, f128mem, loadv2f64, - "movupd", SSEPackedDouble, 0>, TB, OpSize, VEX; + "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>, + TB, OpSize, VEX; defm VMOVAPSY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv8f32, - "movaps", SSEPackedSingle>, TB, VEX; + "movaps", SSEPackedSingle, SSE_MOVA_ITINS>, + TB, VEX; defm VMOVAPDY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv4f64, - "movapd", SSEPackedDouble>, TB, OpSize, VEX; + "movapd", SSEPackedDouble, SSE_MOVA_ITINS>, + TB, OpSize, VEX; defm VMOVUPSY : sse12_mov_packed<0x10, VR256, f256mem, loadv8f32, - "movups", SSEPackedSingle>, TB, VEX; + "movups", SSEPackedSingle, SSE_MOVU_ITINS>, + TB, VEX; defm VMOVUPDY : sse12_mov_packed<0x10, VR256, f256mem, loadv4f64, - "movupd", SSEPackedDouble, 0>, TB, OpSize, VEX; + "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>, + TB, OpSize, VEX; defm MOVAPS : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv4f32, - "movaps", SSEPackedSingle>, TB; + "movaps", SSEPackedSingle, SSE_MOVA_ITINS>, + TB; defm MOVAPD : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv2f64, - "movapd", SSEPackedDouble>, TB, OpSize; + "movapd", SSEPackedDouble, SSE_MOVA_ITINS>, + TB, OpSize; defm MOVUPS : sse12_mov_packed<0x10, VR128, f128mem, loadv4f32, - "movups", SSEPackedSingle>, TB; + "movups", SSEPackedSingle, SSE_MOVU_ITINS>, + TB; defm MOVUPD : sse12_mov_packed<0x10, VR128, f128mem, loadv2f64, - "movupd", SSEPackedDouble, 0>, TB, OpSize; + "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>, + TB, OpSize; def VMOVAPSmr : VPSI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movaps\t{$src, $dst|$dst, $src}", - [(alignedstore (v4f32 VR128:$src), addr:$dst)]>, VEX; + [(alignedstore (v4f32 VR128:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>, VEX; def VMOVAPDmr : VPDI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movapd\t{$src, $dst|$dst, $src}", - [(alignedstore (v2f64 VR128:$src), addr:$dst)]>, VEX; + [(alignedstore (v2f64 VR128:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>, VEX; def VMOVUPSmr : VPSI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movups\t{$src, $dst|$dst, $src}", - [(store (v4f32 VR128:$src), addr:$dst)]>, VEX; + [(store (v4f32 VR128:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>, VEX; def VMOVUPDmr : VPDI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movupd\t{$src, $dst|$dst, $src}", - [(store (v2f64 VR128:$src), addr:$dst)]>, VEX; + [(store (v2f64 VR128:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>, VEX; def VMOVAPSYmr : VPSI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movaps\t{$src, $dst|$dst, $src}", - [(alignedstore256 (v8f32 VR256:$src), addr:$dst)]>, VEX; + [(alignedstore256 (v8f32 VR256:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>, VEX; def VMOVAPDYmr : VPDI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movapd\t{$src, $dst|$dst, $src}", - [(alignedstore256 (v4f64 VR256:$src), addr:$dst)]>, VEX; + [(alignedstore256 (v4f64 VR256:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>, VEX; def VMOVUPSYmr : VPSI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movups\t{$src, $dst|$dst, $src}", - [(store (v8f32 VR256:$src), addr:$dst)]>, VEX; + [(store (v8f32 VR256:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>, VEX; def VMOVUPDYmr : VPDI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movupd\t{$src, $dst|$dst, $src}", - [(store (v4f64 VR256:$src), addr:$dst)]>, VEX; + [(store (v4f64 VR256:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>, VEX; // For disassembler let isCodeGenOnly = 1 in { def VMOVAPSrr_REV : VPSI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movaps\t{$src, $dst|$dst, $src}", []>, VEX; + "movaps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def VMOVAPDrr_REV : VPDI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movapd\t{$src, $dst|$dst, $src}", []>, VEX; + "movapd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def VMOVUPSrr_REV : VPSI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movups\t{$src, $dst|$dst, $src}", []>, VEX; + "movups\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, VEX; def VMOVUPDrr_REV : VPDI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movupd\t{$src, $dst|$dst, $src}", []>, VEX; + "movupd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, VEX; def VMOVAPSYrr_REV : VPSI<0x29, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movaps\t{$src, $dst|$dst, $src}", []>, VEX; + "movaps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def VMOVAPDYrr_REV : VPDI<0x29, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movapd\t{$src, $dst|$dst, $src}", []>, VEX; + "movapd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def VMOVUPSYrr_REV : VPSI<0x11, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movups\t{$src, $dst|$dst, $src}", []>, VEX; + "movups\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, VEX; def VMOVUPDYrr_REV : VPDI<0x11, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movupd\t{$src, $dst|$dst, $src}", []>, VEX; + "movupd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, VEX; } let Predicates = [HasAVX] in { @@ -796,27 +955,35 @@ def : Pat<(int_x86_avx_storeu_pd_256 addr:$dst, VR256:$src), def MOVAPSmr : PSI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movaps\t{$src, $dst|$dst, $src}", - [(alignedstore (v4f32 VR128:$src), addr:$dst)]>; + [(alignedstore (v4f32 VR128:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>; def MOVAPDmr : PDI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movapd\t{$src, $dst|$dst, $src}", - [(alignedstore (v2f64 VR128:$src), addr:$dst)]>; + [(alignedstore (v2f64 VR128:$src), addr:$dst)], + IIC_SSE_MOVA_P_MR>; def MOVUPSmr : PSI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movups\t{$src, $dst|$dst, $src}", - [(store (v4f32 VR128:$src), addr:$dst)]>; + [(store (v4f32 VR128:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>; def MOVUPDmr : PDI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movupd\t{$src, $dst|$dst, $src}", - [(store (v2f64 VR128:$src), addr:$dst)]>; + [(store (v2f64 VR128:$src), addr:$dst)], + IIC_SSE_MOVU_P_MR>; // For disassembler let isCodeGenOnly = 1 in { def MOVAPSrr_REV : PSI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movaps\t{$src, $dst|$dst, $src}", []>; + "movaps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>; def MOVAPDrr_REV : PDI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movapd\t{$src, $dst|$dst, $src}", []>; + "movapd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>; def MOVUPSrr_REV : PSI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movups\t{$src, $dst|$dst, $src}", []>; + "movups\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>; def MOVUPDrr_REV : PDI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movupd\t{$src, $dst|$dst, $src}", []>; + "movupd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>; } let Predicates = [HasAVX] in { @@ -912,13 +1079,17 @@ let Predicates = [HasSSE1] in { // bits are disregarded. FIXME: Set encoding to pseudo! let neverHasSideEffects = 1 in { def FsVMOVAPSrr : VPSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src), - "movaps\t{$src, $dst|$dst, $src}", []>, VEX; + "movaps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def FsVMOVAPDrr : VPDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src), - "movapd\t{$src, $dst|$dst, $src}", []>, VEX; + "movapd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, VEX; def FsMOVAPSrr : PSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src), - "movaps\t{$src, $dst|$dst, $src}", []>; + "movaps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>; def FsMOVAPDrr : PDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src), - "movapd\t{$src, $dst|$dst, $src}", []>; + "movapd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>; } // Alias instruction to load FR32 or FR64 from f128mem using movaps. Upper @@ -927,17 +1098,21 @@ let canFoldAsLoad = 1, isReMaterializable = 1 in { let isCodeGenOnly = 1 in { def FsVMOVAPSrm : VPSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src), "movaps\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (alignedloadfsf32 addr:$src))]>, VEX; + [(set FR32:$dst, (alignedloadfsf32 addr:$src))], + IIC_SSE_MOVA_P_RM>, VEX; def FsVMOVAPDrm : VPDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src), "movapd\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>, VEX; + [(set FR64:$dst, (alignedloadfsf64 addr:$src))], + IIC_SSE_MOVA_P_RM>, VEX; } def FsMOVAPSrm : PSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src), "movaps\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (alignedloadfsf32 addr:$src))]>; + [(set FR32:$dst, (alignedloadfsf32 addr:$src))], + IIC_SSE_MOVA_P_RM>; def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src), "movapd\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>; + [(set FR64:$dst, (alignedloadfsf64 addr:$src))], + IIC_SSE_MOVA_P_RM>; } //===----------------------------------------------------------------------===// @@ -946,48 +1121,54 @@ def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src), multiclass sse12_mov_hilo_packed<bits<8>opc, RegisterClass RC, SDNode psnode, SDNode pdnode, string base_opc, - string asm_opr> { + string asm_opr, InstrItinClass itin> { def PSrm : PI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, f64mem:$src2), !strconcat(base_opc, "s", asm_opr), [(set RC:$dst, (psnode RC:$src1, (bc_v4f32 (v2f64 (scalar_to_vector (loadf64 addr:$src2))))))], - IIC_DEFAULT, SSEPackedSingle>, TB; + itin, SSEPackedSingle>, TB; def PDrm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, f64mem:$src2), !strconcat(base_opc, "d", asm_opr), [(set RC:$dst, (v2f64 (pdnode RC:$src1, (scalar_to_vector (loadf64 addr:$src2)))))], - IIC_DEFAULT, SSEPackedDouble>, TB, OpSize; + itin, SSEPackedDouble>, TB, OpSize; } let AddedComplexity = 20 in { defm VMOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp", - "\t{$src2, $src1, $dst|$dst, $src1, $src2}">, VEX_4V; + "\t{$src2, $src1, $dst|$dst, $src1, $src2}", + IIC_SSE_MOV_LH>, VEX_4V; } let Constraints = "$src1 = $dst", AddedComplexity = 20 in { defm MOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp", - "\t{$src2, $dst|$dst, $src2}">; + "\t{$src2, $dst|$dst, $src2}", + IIC_SSE_MOV_LH>; } def VMOVLPSmr : VPSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movlps\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (bc_v2f64 (v4f32 VR128:$src)), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOV_LH>, VEX; def VMOVLPDmr : VPDI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movlpd\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (v2f64 VR128:$src), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOV_LH>, VEX; def MOVLPSmr : PSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movlps\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (bc_v2f64 (v4f32 VR128:$src)), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOV_LH>; def MOVLPDmr : PDI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movlpd\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (v2f64 VR128:$src), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOV_LH>; let Predicates = [HasAVX] in { // Shuffle with VMOVLPS @@ -1064,11 +1245,13 @@ let Predicates = [HasSSE2] in { let AddedComplexity = 20 in { defm VMOVH : sse12_mov_hilo_packed<0x16, VR128, X86Movlhps, X86Movlhpd, "movhp", - "\t{$src2, $src1, $dst|$dst, $src1, $src2}">, VEX_4V; + "\t{$src2, $src1, $dst|$dst, $src1, $src2}", + IIC_SSE_MOV_LH>, VEX_4V; } let Constraints = "$src1 = $dst", AddedComplexity = 20 in { defm MOVH : sse12_mov_hilo_packed<0x16, VR128, X86Movlhps, X86Movlhpd, "movhp", - "\t{$src2, $dst|$dst, $src2}">; + "\t{$src2, $dst|$dst, $src2}", + IIC_SSE_MOV_LH>; } // v2f64 extract element 1 is always custom lowered to unpack high to low @@ -1078,23 +1261,23 @@ def VMOVHPSmr : VPSI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), [(store (f64 (vector_extract (X86Unpckh (bc_v2f64 (v4f32 VR128:$src)), (bc_v2f64 (v4f32 VR128:$src))), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>, VEX; def VMOVHPDmr : VPDI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movhpd\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (v2f64 (X86Unpckh VR128:$src, VR128:$src)), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>, VEX; def MOVHPSmr : PSI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movhps\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (X86Unpckh (bc_v2f64 (v4f32 VR128:$src)), (bc_v2f64 (v4f32 VR128:$src))), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>; def MOVHPDmr : PDI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src), "movhpd\t{$src, $dst|$dst, $src}", [(store (f64 (vector_extract (v2f64 (X86Unpckh VR128:$src, VR128:$src)), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>; let Predicates = [HasAVX] in { // VMOVHPS patterns @@ -1143,13 +1326,15 @@ let AddedComplexity = 20 in { (ins VR128:$src1, VR128:$src2), "movlhps\t{$src2, $src1, $dst|$dst, $src1, $src2}", [(set VR128:$dst, - (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))]>, + (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))], + IIC_SSE_MOV_LH>, VEX_4V; def VMOVHLPSrr : VPSI<0x12, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2), "movhlps\t{$src2, $src1, $dst|$dst, $src1, $src2}", [(set VR128:$dst, - (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))]>, + (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))], + IIC_SSE_MOV_LH>, VEX_4V; } let Constraints = "$src1 = $dst", AddedComplexity = 20 in { @@ -1157,12 +1342,14 @@ let Constraints = "$src1 = $dst", AddedComplexity = 20 in { (ins VR128:$src1, VR128:$src2), "movlhps\t{$src2, $dst|$dst, $src2}", [(set VR128:$dst, - (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))]>; + (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))], + IIC_SSE_MOV_LH>; def MOVHLPSrr : PSI<0x12, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2), "movhlps\t{$src2, $dst|$dst, $src2}", [(set VR128:$dst, - (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))]>; + (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))], + IIC_SSE_MOV_LH>; } let Predicates = [HasAVX] in { @@ -1193,24 +1380,50 @@ let Predicates = [HasSSE1] in { // SSE 1 & 2 - Conversion Instructions //===----------------------------------------------------------------------===// +def SSE_CVT_PD : OpndItins< + IIC_SSE_CVT_PD_RR, IIC_SSE_CVT_PD_RM +>; + +def SSE_CVT_PS : OpndItins< + IIC_SSE_CVT_PS_RR, IIC_SSE_CVT_PS_RM +>; + +def SSE_CVT_Scalar : OpndItins< + IIC_SSE_CVT_Scalar_RR, IIC_SSE_CVT_Scalar_RM +>; + +def SSE_CVT_SS2SI_32 : OpndItins< + IIC_SSE_CVT_SS2SI32_RR, IIC_SSE_CVT_SS2SI32_RM +>; + +def SSE_CVT_SS2SI_64 : OpndItins< + IIC_SSE_CVT_SS2SI64_RR, IIC_SSE_CVT_SS2SI64_RM +>; + +def SSE_CVT_SD2SI : OpndItins< + IIC_SSE_CVT_SD2SI_RR, IIC_SSE_CVT_SD2SI_RM +>; + multiclass sse12_cvt_s<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC, SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag, - string asm> { + string asm, OpndItins itins> { def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm, - [(set DstRC:$dst, (OpNode SrcRC:$src))]>; + [(set DstRC:$dst, (OpNode SrcRC:$src))], + itins.rr>; def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm, - [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))]>; + [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))], + itins.rm>; } multiclass sse12_cvt_p<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC, SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag, - string asm, Domain d> { + string asm, Domain d, OpndItins itins> { def rr : PI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm, [(set DstRC:$dst, (OpNode SrcRC:$src))], - IIC_DEFAULT, d>; + itins.rr, d>; def rm : PI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm, [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))], - IIC_DEFAULT, d>; + itins.rm, d>; } multiclass sse12_vcvt_avx<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC, @@ -1226,32 +1439,36 @@ let neverHasSideEffects = 1 in { } defm VCVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32, - "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX, - VEX_LIG; + "cvttss2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_32>, + XS, VEX, VEX_LIG; defm VCVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32, - "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX, - VEX_W, VEX_LIG; + "cvttss2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_64>, + XS, VEX, VEX_W, VEX_LIG; defm VCVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64, - "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, VEX, - VEX_LIG; + "cvttsd2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SD2SI>, + XD, VEX, VEX_LIG; defm VCVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64, - "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, - VEX, VEX_W, VEX_LIG; + "cvttsd2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SD2SI>, + XD, VEX, VEX_W, VEX_LIG; // The assembler can recognize rr 64-bit instructions by seeing a rxx // register, but the same isn't true when only using memory operands, // provide other assembly "l" and "q" forms to address this explicitly // where appropriate to do so. -defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">, XS, - VEX_4V, VEX_LIG; -defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">, XS, - VEX_4V, VEX_W, VEX_LIG; -defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">, XD, - VEX_4V, VEX_LIG; -defm VCVTSI2SDL : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd{l}">, XD, - VEX_4V, VEX_LIG; -defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">, XD, - VEX_4V, VEX_W, VEX_LIG; +defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">, + XS, VEX_4V, VEX_LIG; +defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">, + XS, VEX_4V, VEX_W, VEX_LIG; +defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">, + XD, VEX_4V, VEX_LIG; +defm VCVTSI2SDL : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd{l}">, + XD, VEX_4V, VEX_LIG; +defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">, + XD, VEX_4V, VEX_W, VEX_LIG; let Predicates = [HasAVX], AddedComplexity = 1 in { def : Pat<(f32 (sint_to_fp (loadi32 addr:$src))), @@ -1274,137 +1491,163 @@ let Predicates = [HasAVX], AddedComplexity = 1 in { } defm CVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32, - "cvttss2si\t{$src, $dst|$dst, $src}">, XS; + "cvttss2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_32>, XS; defm CVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32, - "cvttss2si{q}\t{$src, $dst|$dst, $src}">, XS, REX_W; + "cvttss2si{q}\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_64>, XS, REX_W; defm CVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64, - "cvttsd2si\t{$src, $dst|$dst, $src}">, XD; + "cvttsd2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SD2SI>, XD; defm CVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64, - "cvttsd2si{q}\t{$src, $dst|$dst, $src}">, XD, REX_W; + "cvttsd2si{q}\t{$src, $dst|$dst, $src}", + SSE_CVT_SD2SI>, XD, REX_W; defm CVTSI2SS : sse12_cvt_s<0x2A, GR32, FR32, sint_to_fp, i32mem, loadi32, - "cvtsi2ss\t{$src, $dst|$dst, $src}">, XS; + "cvtsi2ss\t{$src, $dst|$dst, $src}", + SSE_CVT_Scalar>, XS; defm CVTSI2SS64 : sse12_cvt_s<0x2A, GR64, FR32, sint_to_fp, i64mem, loadi64, - "cvtsi2ss{q}\t{$src, $dst|$dst, $src}">, XS, REX_W; + "cvtsi2ss{q}\t{$src, $dst|$dst, $src}", + SSE_CVT_Scalar>, XS, REX_W; defm CVTSI2SD : sse12_cvt_s<0x2A, GR32, FR64, sint_to_fp, i32mem, loadi32, - "cvtsi2sd\t{$src, $dst|$dst, $src}">, XD; + "cvtsi2sd\t{$src, $dst|$dst, $src}", + SSE_CVT_Scalar>, XD; defm CVTSI2SD64 : sse12_cvt_s<0x2A, GR64, FR64, sint_to_fp, i64mem, loadi64, - "cvtsi2sd{q}\t{$src, $dst|$dst, $src}">, XD, REX_W; + "cvtsi2sd{q}\t{$src, $dst|$dst, $src}", + SSE_CVT_Scalar>, XD, REX_W; // Conversion Instructions Intrinsics - Match intrinsics which expect MM // and/or XMM operand(s). multiclass sse12_cvt_sint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC, Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag, - string asm> { + string asm, OpndItins itins> { def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), !strconcat(asm, "\t{$src, $dst|$dst, $src}"), - [(set DstRC:$dst, (Int SrcRC:$src))]>; + [(set DstRC:$dst, (Int SrcRC:$src))], itins.rr>; def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), !strconcat(asm, "\t{$src, $dst|$dst, $src}"), - [(set DstRC:$dst, (Int (ld_frag addr:$src)))]>; + [(set DstRC:$dst, (Int (ld_frag addr:$src)))], itins.rm>; } multiclass sse12_cvt_sint_3addr<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC, Intrinsic Int, X86MemOperand x86memop, - PatFrag ld_frag, string asm, bit Is2Addr = 1> { + PatFrag ld_frag, string asm, OpndItins itins, + bit Is2Addr = 1> { def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins DstRC:$src1, SrcRC:$src2), !if(Is2Addr, !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"), !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))]>; + [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))], + itins.rr>; def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins DstRC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"), !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))]>; + [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))], + itins.rm>; } defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si, - f128mem, load, "cvtsd2si">, XD, VEX, VEX_LIG; + f128mem, load, "cvtsd2si", SSE_CVT_SD2SI>, XD, VEX, VEX_LIG; defm VCVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, - int_x86_sse2_cvtsd2si64, f128mem, load, "cvtsd2si">, - XD, VEX, VEX_W, VEX_LIG; + int_x86_sse2_cvtsd2si64, f128mem, load, "cvtsd2si", + SSE_CVT_SD2SI>, XD, VEX, VEX_W, VEX_LIG; defm CVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si, - f128mem, load, "cvtsd2si{l}">, XD; + f128mem, load, "cvtsd2si{l}", SSE_CVT_SD2SI>, XD; defm CVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse2_cvtsd2si64, - f128mem, load, "cvtsd2si{q}">, XD, REX_W; + f128mem, load, "cvtsd2si{q}", SSE_CVT_SD2SI>, XD, REX_W; defm Int_VCVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128, - int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss", 0>, XS, VEX_4V; + int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss", + SSE_CVT_Scalar, 0>, XS, VEX_4V; defm Int_VCVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128, - int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss", 0>, XS, VEX_4V, + int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss", + SSE_CVT_Scalar, 0>, XS, VEX_4V, VEX_W; defm Int_VCVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128, - int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd", 0>, XD, VEX_4V; + int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd", + SSE_CVT_Scalar, 0>, XD, VEX_4V; defm Int_VCVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128, - int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd", 0>, XD, + int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd", + SSE_CVT_Scalar, 0>, XD, VEX_4V, VEX_W; let Constraints = "$src1 = $dst" in { defm Int_CVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128, int_x86_sse_cvtsi2ss, i32mem, loadi32, - "cvtsi2ss">, XS; + "cvtsi2ss", SSE_CVT_Scalar>, XS; defm Int_CVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128, int_x86_sse_cvtsi642ss, i64mem, loadi64, - "cvtsi2ss{q}">, XS, REX_W; + "cvtsi2ss{q}", SSE_CVT_Scalar>, XS, REX_W; defm Int_CVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128, int_x86_sse2_cvtsi2sd, i32mem, loadi32, - "cvtsi2sd">, XD; + "cvtsi2sd", SSE_CVT_Scalar>, XD; defm Int_CVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128, int_x86_sse2_cvtsi642sd, i64mem, loadi64, - "cvtsi2sd">, XD, REX_W; + "cvtsi2sd", SSE_CVT_Scalar>, XD, REX_W; } /// SSE 1 Only // Aliases for intrinsics defm Int_VCVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si, - f32mem, load, "cvttss2si">, XS, VEX; + f32mem, load, "cvttss2si", + SSE_CVT_SS2SI_32>, XS, VEX; defm Int_VCVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, int_x86_sse_cvttss2si64, f32mem, load, - "cvttss2si">, XS, VEX, VEX_W; + "cvttss2si", SSE_CVT_SS2SI_64>, + XS, VEX, VEX_W; defm Int_VCVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si, - f128mem, load, "cvttsd2si">, XD, VEX; + f128mem, load, "cvttsd2si", SSE_CVT_SD2SI>, + XD, VEX; defm Int_VCVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, int_x86_sse2_cvttsd2si64, f128mem, load, - "cvttsd2si">, XD, VEX, VEX_W; + "cvttsd2si", SSE_CVT_SD2SI>, + XD, VEX, VEX_W; defm Int_CVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si, - f32mem, load, "cvttss2si">, XS; + f32mem, load, "cvttss2si", + SSE_CVT_SS2SI_32>, XS; defm Int_CVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, int_x86_sse_cvttss2si64, f32mem, load, - "cvttss2si{q}">, XS, REX_W; + "cvttss2si{q}", SSE_CVT_SS2SI_64>, + XS, REX_W; defm Int_CVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si, - f128mem, load, "cvttsd2si">, XD; + f128mem, load, "cvttsd2si", SSE_CVT_SD2SI>, + XD; defm Int_CVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, int_x86_sse2_cvttsd2si64, f128mem, load, - "cvttsd2si{q}">, XD, REX_W; + "cvttsd2si{q}", SSE_CVT_SD2SI>, + XD, REX_W; let Pattern = []<dag> in { defm VCVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load, - "cvtss2si{l}\t{$src, $dst|$dst, $src}">, XS, - VEX, VEX_LIG; + "cvtss2si{l}\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_32>, XS, VEX, VEX_LIG; defm VCVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load, - "cvtss2si\t{$src, $dst|$dst, $src}">, XS, VEX, - VEX_W, VEX_LIG; + "cvtss2si\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_64>, XS, VEX, VEX_W, VEX_LIG; defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load, "cvtdq2ps\t{$src, $dst|$dst, $src}", - SSEPackedSingle>, TB, VEX; + SSEPackedSingle, SSE_CVT_PS>, TB, VEX; defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, undef, i256mem, load, "cvtdq2ps\t{$src, $dst|$dst, $src}", - SSEPackedSingle>, TB, VEX; + SSEPackedSingle, SSE_CVT_PS>, TB, VEX; } let Pattern = []<dag> in { defm CVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load /*dummy*/, - "cvtss2si{l}\t{$src, $dst|$dst, $src}">, XS; + "cvtss2si{l}\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_32>, XS; defm CVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load /*dummy*/, - "cvtss2si{q}\t{$src, $dst|$dst, $src}">, XS, REX_W; + "cvtss2si{q}\t{$src, $dst|$dst, $src}", + SSE_CVT_SS2SI_64>, XS, REX_W; defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load /*dummy*/, "cvtdq2ps\t{$src, $dst|$dst, $src}", - SSEPackedSingle>, TB; /* PD SSE3 form is avaiable */ + SSEPackedSingle, SSE_CVT_PS>, + TB; /* PD SSE3 form is avaiable */ } let Predicates = [HasAVX] in { @@ -1434,43 +1677,51 @@ let Predicates = [HasSSE1] in { // Convert scalar double to scalar single def VCVTSD2SSrr : VSDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src1, FR64:$src2), - "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, - VEX_4V, VEX_LIG; + "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [], + IIC_SSE_CVT_Scalar_RR>, VEX_4V, VEX_LIG; let mayLoad = 1 in def VCVTSD2SSrm : I<0x5A, MRMSrcMem, (outs FR32:$dst), (ins FR64:$src1, f64mem:$src2), "vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", - []>, XD, Requires<[HasAVX, OptForSize]>, VEX_4V, VEX_LIG; + [], IIC_SSE_CVT_Scalar_RM>, + XD, Requires<[HasAVX, OptForSize]>, VEX_4V, VEX_LIG; def : Pat<(f32 (fround FR64:$src)), (VCVTSD2SSrr FR64:$src, FR64:$src)>, Requires<[HasAVX]>; def CVTSD2SSrr : SDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src), "cvtsd2ss\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (fround FR64:$src))]>; + [(set FR32:$dst, (fround FR64:$src))], + IIC_SSE_CVT_Scalar_RR>; def CVTSD2SSrm : I<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src), "cvtsd2ss\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (fround (loadf64 addr:$src)))]>, XD, + [(set FR32:$dst, (fround (loadf64 addr:$src)))], + IIC_SSE_CVT_Scalar_RM>, + XD, Requires<[HasSSE2, OptForSize]>; defm Int_VCVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128, - int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss", 0>, + int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss", + SSE_CVT_Scalar, 0>, XS, VEX_4V; let Constraints = "$src1 = $dst" in defm Int_CVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128, - int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss">, XS; + int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss", + SSE_CVT_Scalar>, XS; // Convert scalar single to scalar double // SSE2 instructions with XS prefix def VCVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src1, FR32:$src2), "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", - []>, XS, Requires<[HasAVX]>, VEX_4V, VEX_LIG; + [], IIC_SSE_CVT_Scalar_RR>, + XS, Requires<[HasAVX]>, VEX_4V, VEX_LIG; let mayLoad = 1 in def VCVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins FR32:$src1, f32mem:$src2), "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", - []>, XS, VEX_4V, VEX_LIG, Requires<[HasAVX, OptForSize]>; + [], IIC_SSE_CVT_Scalar_RM>, + XS, VEX_4V, VEX_LIG, Requires<[HasAVX, OptForSize]>; let Predicates = [HasAVX] in { def : Pat<(f64 (fextend FR32:$src)), @@ -1487,11 +1738,13 @@ def : Pat<(extloadf32 addr:$src), def CVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src), "cvtss2sd\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (fextend FR32:$src))]>, XS, + [(set FR64:$dst, (fextend FR32:$src))], + IIC_SSE_CVT_Scalar_RR>, XS, Requires<[HasSSE2]>; def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src), "cvtss2sd\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (extloadf32 addr:$src))]>, XS, + [(set FR64:$dst, (extloadf32 addr:$src))], + IIC_SSE_CVT_Scalar_RM>, XS, Requires<[HasSSE2, OptForSize]>; // extload f32 -> f64. This matches load+fextend because we have a hack in @@ -1508,26 +1761,30 @@ def Int_VCVTSS2SDrr: I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2), "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1, - VR128:$src2))]>, XS, VEX_4V, + VR128:$src2))], + IIC_SSE_CVT_Scalar_RR>, XS, VEX_4V, Requires<[HasAVX]>; def Int_VCVTSS2SDrm: I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, f32mem:$src2), "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1, - (load addr:$src2)))]>, XS, VEX_4V, + (load addr:$src2)))], + IIC_SSE_CVT_Scalar_RM>, XS, VEX_4V, Requires<[HasAVX]>; let Constraints = "$src1 = $dst" in { // SSE2 instructions with XS prefix def Int_CVTSS2SDrr: I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2), "cvtss2sd\t{$src2, $dst|$dst, $src2}", [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1, - VR128:$src2))]>, XS, + VR128:$src2))], + IIC_SSE_CVT_Scalar_RR>, XS, Requires<[HasSSE2]>; def Int_CVTSS2SDrm: I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, f32mem:$src2), "cvtss2sd\t{$src2, $dst|$dst, $src2}", [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1, - (load addr:$src2)))]>, XS, + (load addr:$src2)))], + IIC_SSE_CVT_Scalar_RM>, XS, Requires<[HasSSE2]>; } @@ -1535,94 +1792,116 @@ def Int_CVTSS2SDrm: I<0x5A, MRMSrcMem, // SSE2 instructions without OpSize prefix def Int_VCVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vcvtdq2ps\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))], + IIC_SSE_CVT_PS_RR>, TB, VEX, Requires<[HasAVX]>; def Int_VCVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "vcvtdq2ps\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtdq2ps - (bitconvert (memopv2i64 addr:$src))))]>, + (bitconvert (memopv2i64 addr:$src))))], + IIC_SSE_CVT_PS_RM>, TB, VEX, Requires<[HasAVX]>; def Int_CVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtdq2ps\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))], + IIC_SSE_CVT_PS_RR>, TB, Requires<[HasSSE2]>; def Int_CVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "cvtdq2ps\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtdq2ps - (bitconvert (memopv2i64 addr:$src))))]>, + (bitconvert (memopv2i64 addr:$src))))], + IIC_SSE_CVT_PS_RM>, TB, Requires<[HasSSE2]>; // FIXME: why the non-intrinsic version is described as SSE3? // SSE2 instructions with XS prefix def Int_VCVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vcvtdq2pd\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))], + IIC_SSE_CVT_PD_RR>, XS, VEX, Requires<[HasAVX]>; def Int_VCVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), "vcvtdq2pd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtdq2pd - (bitconvert (memopv2i64 addr:$src))))]>, + (bitconvert (memopv2i64 addr:$src))))], + IIC_SSE_CVT_PD_RM>, XS, VEX, Requires<[HasAVX]>; def Int_CVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtdq2pd\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))], + IIC_SSE_CVT_PD_RR>, XS, Requires<[HasSSE2]>; def Int_CVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), "cvtdq2pd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtdq2pd - (bitconvert (memopv2i64 addr:$src))))]>, + (bitconvert (memopv2i64 addr:$src))))], + IIC_SSE_CVT_PD_RM>, XS, Requires<[HasSSE2]>; // Convert packed single/double fp to doubleword def VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RR>, VEX; def VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RM>, VEX; def VCVTPS2DQYrr : VPDI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RR>, VEX; def VCVTPS2DQYrm : VPDI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RM>, VEX; def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RR>; def CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtps2dq\t{$src, $dst|$dst, $src}", []>; + "cvtps2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PS_RM>; def Int_VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtps2dq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))], + IIC_SSE_CVT_PS_RR>, VEX; def Int_VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvtps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtps2dq - (memop addr:$src)))]>, VEX; + (memop addr:$src)))], + IIC_SSE_CVT_PS_RM>, VEX; def Int_CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtps2dq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))]>; + [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))], + IIC_SSE_CVT_PS_RR>; def Int_CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvtps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtps2dq - (memop addr:$src)))]>; + (memop addr:$src)))], + IIC_SSE_CVT_PS_RM>; // SSE2 packed instructions with XD prefix def Int_VCVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vcvtpd2dq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))], + IIC_SSE_CVT_PD_RR>, XD, VEX, Requires<[HasAVX]>; def Int_VCVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "vcvtpd2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtpd2dq - (memop addr:$src)))]>, + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>, XD, VEX, Requires<[HasAVX]>; def Int_CVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtpd2dq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))], + IIC_SSE_CVT_PD_RR>, XD, Requires<[HasSSE2]>; def Int_CVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvtpd2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtpd2dq - (memop addr:$src)))]>, + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>, XD, Requires<[HasSSE2]>; @@ -1631,28 +1910,34 @@ def Int_CVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), def VCVTTPS2DQrr : VSSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (int_x86_sse2_cvttps2dq VR128:$src))]>, VEX; + (int_x86_sse2_cvttps2dq VR128:$src))], + IIC_SSE_CVT_PS_RR>, VEX; def VCVTTPS2DQrm : VSSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvttps2dq - (memop addr:$src)))]>, VEX; + (memop addr:$src)))], + IIC_SSE_CVT_PS_RM>, VEX; def VCVTTPS2DQYrr : VSSI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR256:$dst, - (int_x86_avx_cvtt_ps2dq_256 VR256:$src))]>, VEX; + (int_x86_avx_cvtt_ps2dq_256 VR256:$src))], + IIC_SSE_CVT_PS_RR>, VEX; def VCVTTPS2DQYrm : VSSI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256 - (memopv8f32 addr:$src)))]>, VEX; + (memopv8f32 addr:$src)))], + IIC_SSE_CVT_PS_RM>, VEX; def CVTTPS2DQrr : SSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (int_x86_sse2_cvttps2dq VR128:$src))]>; + (int_x86_sse2_cvttps2dq VR128:$src))], + IIC_SSE_CVT_PS_RR>; def CVTTPS2DQrm : SSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvttps2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (int_x86_sse2_cvttps2dq (memop addr:$src)))]>; + (int_x86_sse2_cvttps2dq (memop addr:$src)))], + IIC_SSE_CVT_PS_RM>; let Predicates = [HasAVX] in { def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))), @@ -1691,72 +1976,91 @@ let Predicates = [HasSSE2] in { def VCVTTPD2DQrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvttpd2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (int_x86_sse2_cvttpd2dq VR128:$src))]>, VEX; + (int_x86_sse2_cvttpd2dq VR128:$src))], + IIC_SSE_CVT_PD_RR>, VEX; let isCodeGenOnly = 1 in def VCVTTPD2DQrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvttpd2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvttpd2dq - (memop addr:$src)))]>, VEX; + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>, VEX; def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvttpd2dq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvttpd2dq VR128:$src))]>; + [(set VR128:$dst, (int_x86_sse2_cvttpd2dq VR128:$src))], + IIC_SSE_CVT_PD_RR>; def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src), "cvttpd2dq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvttpd2dq - (memop addr:$src)))]>; + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>; // The assembler can recognize rr 256-bit instructions by seeing a ymm // register, but the same isn't true when using memory operands instead. // Provide other assembly rr and rm forms to address this explicitly. def VCVTTPD2DQXrYr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src), - "cvttpd2dq\t{$src, $dst|$dst, $src}", []>, VEX; + "cvttpd2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; // XMM only def VCVTTPD2DQXrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvttpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX; + "cvttpd2dqx\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; def VCVTTPD2DQXrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvttpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX; + "cvttpd2dqx\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, VEX; // YMM only def VCVTTPD2DQYrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src), - "cvttpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX; + "cvttpd2dqy\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src), - "cvttpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L; + "cvttpd2dqy\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, VEX, VEX_L; // Convert packed single to packed double let Predicates = [HasAVX] in { // SSE2 instructions without OpSize prefix def VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX; + "vcvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, TB, VEX; def VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), - "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX; + "vcvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, TB, VEX; def VCVTPS2PDYrr : I<0x5A, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src), - "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX; + "vcvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, TB, VEX; def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src), - "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX; + "vcvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, TB, VEX; } def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtps2pd\t{$src, $dst|$dst, $src}", []>, TB; + "cvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, TB; def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), - "cvtps2pd\t{$src, $dst|$dst, $src}", []>, TB; + "cvtps2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, TB; def Int_VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vcvtps2pd\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))], + IIC_SSE_CVT_PD_RR>, TB, VEX, Requires<[HasAVX]>; def Int_VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), "vcvtps2pd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtps2pd - (load addr:$src)))]>, + (load addr:$src)))], + IIC_SSE_CVT_PD_RM>, TB, VEX, Requires<[HasAVX]>; def Int_CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtps2pd\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))]>, + [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))], + IIC_SSE_CVT_PD_RR>, TB, Requires<[HasSSE2]>; def Int_CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), "cvtps2pd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtps2pd - (load addr:$src)))]>, + (load addr:$src)))], + IIC_SSE_CVT_PD_RM>, TB, Requires<[HasSSE2]>; // Convert packed double to packed single @@ -1764,42 +2068,54 @@ def Int_CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), // register, but the same isn't true when using memory operands instead. // Provide other assembly rr and rm forms to address this explicitly. def VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtpd2ps\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtpd2ps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; def VCVTPD2PSXrYr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src), - "cvtpd2ps\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtpd2ps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; // XMM only def VCVTPD2PSXrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtpd2psx\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtpd2psx\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; def VCVTPD2PSXrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtpd2psx\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtpd2psx\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, VEX; // YMM only def VCVTPD2PSYrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src), - "cvtpd2psy\t{$src, $dst|$dst, $src}", []>, VEX; + "cvtpd2psy\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>, VEX; def VCVTPD2PSYrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src), - "cvtpd2psy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L; + "cvtpd2psy\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>, VEX, VEX_L; def CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtpd2ps\t{$src, $dst|$dst, $src}", []>; + "cvtpd2ps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>; def CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtpd2ps\t{$src, $dst|$dst, $src}", []>; + "cvtpd2ps\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>; def Int_VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtpd2ps\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))]>; + [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))], + IIC_SSE_CVT_PD_RR>; def Int_VCVTPD2PSrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvtpd2ps\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtpd2ps - (memop addr:$src)))]>; + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>; def Int_CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "cvtpd2ps\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))]>; + [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))], + IIC_SSE_CVT_PD_RR>; def Int_CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), "cvtpd2ps\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (int_x86_sse2_cvtpd2ps - (memop addr:$src)))]>; + (memop addr:$src)))], + IIC_SSE_CVT_PD_RM>; // AVX 256-bit register conversion intrinsics // FIXME: Migrate SSE conversion intrinsics matching to use patterns as below @@ -1847,69 +2163,84 @@ def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))), // sse12_cmp_scalar - sse 1 & 2 compare scalar instructions multiclass sse12_cmp_scalar<RegisterClass RC, X86MemOperand x86memop, SDNode OpNode, ValueType VT, PatFrag ld_frag, - string asm, string asm_alt> { + string asm, string asm_alt, + OpndItins itins> { def rr : SIi8<0xC2, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2, SSECC:$cc), asm, - [(set RC:$dst, (OpNode (VT RC:$src1), RC:$src2, imm:$cc))]>; + [(set RC:$dst, (OpNode (VT RC:$src1), RC:$src2, imm:$cc))], + itins.rr>; def rm : SIi8<0xC2, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2, SSECC:$cc), asm, [(set RC:$dst, (OpNode (VT RC:$src1), - (ld_frag addr:$src2), imm:$cc))]>; + (ld_frag addr:$src2), imm:$cc))], + itins.rm>; // Accept explicit immediate argument form instead of comparison code. let neverHasSideEffects = 1 in { def rr_alt : SIi8<0xC2, MRMSrcReg, (outs RC:$dst), - (ins RC:$src1, RC:$src2, i8imm:$cc), asm_alt, []>; + (ins RC:$src1, RC:$src2, i8imm:$cc), asm_alt, [], + IIC_SSE_ALU_F32S_RR>; let mayLoad = 1 in def rm_alt : SIi8<0xC2, MRMSrcMem, (outs RC:$dst), - (ins RC:$src1, x86memop:$src2, i8imm:$cc), asm_alt, []>; + (ins RC:$src1, x86memop:$src2, i8imm:$cc), asm_alt, [], + IIC_SSE_ALU_F32S_RM>; } } defm VCMPSS : sse12_cmp_scalar<FR32, f32mem, X86cmpss, f32, loadf32, "cmp${cc}ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", - "cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}">, + "cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}", + SSE_ALU_F32S>, XS, VEX_4V, VEX_LIG; defm VCMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmpsd, f64, loadf64, "cmp${cc}sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", - "cmpsd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}">, + "cmpsd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}", + SSE_ALU_F32S>, // same latency as 32 bit compare XD, VEX_4V, VEX_LIG; let Constraints = "$src1 = $dst" in { defm CMPSS : sse12_cmp_scalar<FR32, f32mem, X86cmpss, f32, loadf32, "cmp${cc}ss\t{$src2, $dst|$dst, $src2}", - "cmpss\t{$cc, $src2, $dst|$dst, $src2, $cc}">, + "cmpss\t{$cc, $src2, $dst|$dst, $src2, $cc}", SSE_ALU_F32S>, XS; defm CMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmpsd, f64, loadf64, "cmp${cc}sd\t{$src2, $dst|$dst, $src2}", - "cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}">, + "cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}", + SSE_ALU_F32S>, // same latency as 32 bit compare XD; } multiclass sse12_cmp_scalar_int<RegisterClass RC, X86MemOperand x86memop, - Intrinsic Int, string asm> { + Intrinsic Int, string asm, OpndItins itins> { def rr : SIi8<0xC2, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc), asm, [(set VR128:$dst, (Int VR128:$src1, - VR128:$src, imm:$cc))]>; + VR128:$src, imm:$cc))], + itins.rr>; def rm : SIi8<0xC2, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, x86memop:$src, SSECC:$cc), asm, [(set VR128:$dst, (Int VR128:$src1, - (load addr:$src), imm:$cc))]>; + (load addr:$src), imm:$cc))], + itins.rm>; } // Aliases to match intrinsics which expect XMM operand(s). defm Int_VCMPSS : sse12_cmp_scalar_int<VR128, f32mem, int_x86_sse_cmp_ss, - "cmp${cc}ss\t{$src, $src1, $dst|$dst, $src1, $src}">, + "cmp${cc}ss\t{$src, $src1, $dst|$dst, $src1, $src}", + SSE_ALU_F32S>, XS, VEX_4V; defm Int_VCMPSD : sse12_cmp_scalar_int<VR128, f64mem, int_x86_sse2_cmp_sd, - "cmp${cc}sd\t{$src, $src1, $dst|$dst, $src1, $src}">, + "cmp${cc}sd\t{$src, $src1, $dst|$dst, $src1, $src}", + SSE_ALU_F32S>, // same latency as f32 XD, VEX_4V; let Constraints = "$src1 = $dst" in { defm Int_CMPSS : sse12_cmp_scalar_int<VR128, f32mem, int_x86_sse_cmp_ss, - "cmp${cc}ss\t{$src, $dst|$dst, $src}">, XS; + "cmp${cc}ss\t{$src, $dst|$dst, $src}", + SSE_ALU_F32S>, XS; defm Int_CMPSD : sse12_cmp_scalar_int<VR128, f64mem, int_x86_sse2_cmp_sd, - "cmp${cc}sd\t{$src, $dst|$dst, $src}">, XD; + "cmp${cc}sd\t{$src, $dst|$dst, $src}", + SSE_ALU_F32S>, // same latency as f32 + XD; } @@ -1920,12 +2251,12 @@ multiclass sse12_ord_cmp<bits<8> opc, RegisterClass RC, SDNode OpNode, def rr: PI<opc, MRMSrcReg, (outs), (ins RC:$src1, RC:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"), [(set EFLAGS, (OpNode (vt RC:$src1), RC:$src2))], - IIC_DEFAULT, d>; + IIC_SSE_COMIS_RR, d>; def rm: PI<opc, MRMSrcMem, (outs), (ins RC:$src1, x86memop:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"), [(set EFLAGS, (OpNode (vt RC:$src1), (ld_frag addr:$src2)))], - IIC_DEFAULT, d>; + IIC_SSE_COMIS_RM, d>; } let Defs = [EFLAGS] in { @@ -1983,20 +2314,20 @@ multiclass sse12_cmp_packed<RegisterClass RC, X86MemOperand x86memop, def rri : PIi8<0xC2, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2, SSECC:$cc), asm, [(set RC:$dst, (Int RC:$src1, RC:$src2, imm:$cc))], - IIC_DEFAULT, d>; + IIC_SSE_CMPP_RR, d>; def rmi : PIi8<0xC2, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2, SSECC:$cc), asm, [(set RC:$dst, (Int RC:$src1, (memop addr:$src2), imm:$cc))], - IIC_DEFAULT, d>; + IIC_SSE_CMPP_RM, d>; } // Accept explicit immediate argument form instead of comparison code. def rri_alt : PIi8<0xC2, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2, i8imm:$cc), - asm_alt, [], IIC_DEFAULT, d>; + asm_alt, [], IIC_SSE_CMPP_RR, d>; def rmi_alt : PIi8<0xC2, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$cc), - asm_alt, [], IIC_DEFAULT, d>; + asm_alt, [], IIC_SSE_CMPP_RM, d>; } defm VCMPPS : sse12_cmp_packed<VR128, f128mem, int_x86_sse_cmp_ps, @@ -2071,12 +2402,12 @@ multiclass sse12_shuffle<RegisterClass RC, X86MemOperand x86memop, def rmi : PIi8<0xC6, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$src3), asm, [(set RC:$dst, (vt (X86Shufp RC:$src1, (mem_frag addr:$src2), - (i8 imm:$src3))))], IIC_DEFAULT, d>; + (i8 imm:$src3))))], IIC_SSE_SHUFP, d>; let isConvertibleToThreeAddress = IsConvertibleToThreeAddress in def rri : PIi8<0xC6, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2, i8imm:$src3), asm, [(set RC:$dst, (vt (X86Shufp RC:$src1, RC:$src2, - (i8 imm:$src3))))], IIC_DEFAULT, d>; + (i8 imm:$src3))))], IIC_SSE_SHUFP, d>; } defm VSHUFPS : sse12_shuffle<VR128, f128mem, v4f32, @@ -2160,13 +2491,13 @@ multiclass sse12_unpack_interleave<bits<8> opc, SDNode OpNode, ValueType vt, (outs RC:$dst), (ins RC:$src1, RC:$src2), asm, [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], - IIC_DEFAULT, d>; + IIC_SSE_UNPCK, d>; def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), asm, [(set RC:$dst, (vt (OpNode RC:$src1, (mem_frag addr:$src2))))], - IIC_DEFAULT, d>; + IIC_SSE_UNPCK, d>; } defm VUNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, memopv4f32, @@ -2237,10 +2568,10 @@ multiclass sse12_extr_sign_mask<RegisterClass RC, Intrinsic Int, string asm, Domain d> { def rr32 : PI<0x50, MRMSrcReg, (outs GR32:$dst), (ins RC:$src), !strconcat(asm, "\t{$src, $dst|$dst, $src}"), - [(set GR32:$dst, (Int RC:$src))], IIC_DEFAULT, d>; + [(set GR32:$dst, (Int RC:$src))], IIC_SSE_MOVMSK, d>; def rr64 : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins RC:$src), !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], - IIC_DEFAULT, d>, REX_W; + IIC_SSE_MOVMSK, d>, REX_W; } let Predicates = [HasAVX] in { @@ -2270,17 +2601,17 @@ let Predicates = [HasAVX] in { // Assembler Only def VMOVMSKPSr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src), - "movmskps\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT, + "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK, SSEPackedSingle>, TB, VEX; def VMOVMSKPDr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src), - "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT, + "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK, SSEPackedDouble>, TB, OpSize, VEX; def VMOVMSKPSYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src), - "movmskps\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT, + "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK, SSEPackedSingle>, TB, VEX; def VMOVMSKPDYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src), - "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT, + "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK, SSEPackedDouble>, TB, OpSize, VEX; } @@ -2312,7 +2643,9 @@ let ExeDomain = SSEPackedInt in { // SSE integer instructions /// PDI_binop_rm - Simple SSE2 binary operator. multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode, ValueType OpVT, RegisterClass RC, PatFrag memop_frag, - X86MemOperand x86memop, bit IsCommutable = 0, + X86MemOperand x86memop, + OpndItins itins, + bit IsCommutable = 0, bit Is2Addr = 1> { let isCommutable = IsCommutable in def rr : PDI<opc, MRMSrcReg, (outs RC:$dst), @@ -2320,14 +2653,15 @@ multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode, !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))]>; + [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))], itins.rr>; def rm : PDI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (OpVT (OpNode RC:$src1, - (bitconvert (memop_frag addr:$src2)))))]>; + (bitconvert (memop_frag addr:$src2)))))], + itins.rm>; } } // ExeDomain = SSEPackedInt @@ -2335,35 +2669,35 @@ multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode, let Predicates = [HasAVX] in { defm VPAND : PDI_binop_rm<0xDB, "vpand", and, v2i64, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPOR : PDI_binop_rm<0xEB, "vpor" , or, v2i64, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPXOR : PDI_binop_rm<0xEF, "vpxor", xor, v2i64, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPANDN : PDI_binop_rm<0xDF, "vpandn", X86andnp, v2i64, VR128, memopv2i64, - i128mem, 0, 0>, VEX_4V; + i128mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V; } let Constraints = "$src1 = $dst" in { defm PAND : PDI_binop_rm<0xDB, "pand", and, v2i64, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_BIT_ITINS_P, 1>; defm POR : PDI_binop_rm<0xEB, "por" , or, v2i64, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_BIT_ITINS_P, 1>; defm PXOR : PDI_binop_rm<0xEF, "pxor", xor, v2i64, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_BIT_ITINS_P, 1>; defm PANDN : PDI_binop_rm<0xDF, "pandn", X86andnp, v2i64, VR128, memopv2i64, - i128mem, 0>; + i128mem, SSE_BIT_ITINS_P, 0>; } // Constraints = "$src1 = $dst" let Predicates = [HasAVX2] in { defm VPANDY : PDI_binop_rm<0xDB, "vpand", and, v4i64, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPORY : PDI_binop_rm<0xEB, "vpor", or, v4i64, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPXORY : PDI_binop_rm<0xEF, "vpxor", xor, v4i64, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V; defm VPANDNY : PDI_binop_rm<0xDF, "vpandn", X86andnp, v4i64, VR256, memopv4i64, - i256mem, 0, 0>, VEX_4V; + i256mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V; } //===----------------------------------------------------------------------===// @@ -2373,31 +2707,39 @@ defm VPANDNY : PDI_binop_rm<0xDF, "vpandn", X86andnp, v4i64, VR256, memopv4i64, /// sse12_fp_alias_pack_logical - SSE 1 & 2 aliased packed FP logical ops /// multiclass sse12_fp_alias_pack_logical<bits<8> opc, string OpcodeStr, - SDNode OpNode> { + SDNode OpNode, OpndItins itins> { defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, - FR32, f32, f128mem, memopfsf32, SSEPackedSingle, 0>, TB, VEX_4V; + FR32, f32, f128mem, memopfsf32, SSEPackedSingle, itins, 0>, + TB, VEX_4V; defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, - FR64, f64, f128mem, memopfsf64, SSEPackedDouble, 0>, TB, OpSize, VEX_4V; + FR64, f64, f128mem, memopfsf64, SSEPackedDouble, itins, 0>, + TB, OpSize, VEX_4V; let Constraints = "$src1 = $dst" in { defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, FR32, - f32, f128mem, memopfsf32, SSEPackedSingle>, TB; + f32, f128mem, memopfsf32, SSEPackedSingle, itins>, + TB; defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, FR64, - f64, f128mem, memopfsf64, SSEPackedDouble>, TB, OpSize; + f64, f128mem, memopfsf64, SSEPackedDouble, itins>, + TB, OpSize; } } // Alias bitwise logical operations using SSE logical ops on packed FP values. let mayLoad = 0 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, + SSE_BIT_ITINS_P>; + defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for, + SSE_BIT_ITINS_P>; + defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor, + SSE_BIT_ITINS_P>; } let neverHasSideEffects = 1, Pattern = []<dag>, isCommutable = 0 in - defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef>; + defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef, + SSE_BIT_ITINS_P>; /// sse12_fp_packed_logical - SSE 1 & 2 packed FP logical ops /// @@ -2484,118 +2826,145 @@ let isCommutable = 0 in /// FIXME: once all 256-bit intrinsics are matched, cleanup and refactor those /// classes below multiclass basic_sse12_fp_binop_s<bits<8> opc, string OpcodeStr, SDNode OpNode, + SizeItins itins, bit Is2Addr = 1> { defm SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"), - OpNode, FR32, f32mem, Is2Addr>, XS; + OpNode, FR32, f32mem, + itins.s, Is2Addr>, XS; defm SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"), - OpNode, FR64, f64mem, Is2Addr>, XD; + OpNode, FR64, f64mem, + itins.d, Is2Addr>, XD; } multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr, SDNode OpNode, + SizeItins itins, bit Is2Addr = 1> { let mayLoad = 0 in { defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR128, - v4f32, f128mem, memopv4f32, SSEPackedSingle, Is2Addr>, TB; + v4f32, f128mem, memopv4f32, SSEPackedSingle, itins.s, Is2Addr>, + TB; defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR128, - v2f64, f128mem, memopv2f64, SSEPackedDouble, Is2Addr>, TB, OpSize; + v2f64, f128mem, memopv2f64, SSEPackedDouble, itins.d, Is2Addr>, + TB, OpSize; } } multiclass basic_sse12_fp_binop_p_y<bits<8> opc, string OpcodeStr, - SDNode OpNode> { + SDNode OpNode, + SizeItins itins> { let mayLoad = 0 in { defm PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR256, - v8f32, f256mem, memopv8f32, SSEPackedSingle, 0>, TB; + v8f32, f256mem, memopv8f32, SSEPackedSingle, itins.s, 0>, + TB; defm PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR256, - v4f64, f256mem, memopv4f64, SSEPackedDouble, 0>, TB, OpSize; + v4f64, f256mem, memopv4f64, SSEPackedDouble, itins.d, 0>, + TB, OpSize; } } multiclass basic_sse12_fp_binop_s_int<bits<8> opc, string OpcodeStr, + SizeItins itins, bit Is2Addr = 1> { defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128, - !strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32, Is2Addr>, XS; + !strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32, + itins.s, Is2Addr>, XS; defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128, - !strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64, Is2Addr>, XD; + !strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64, + itins.d, Is2Addr>, XD; } multiclass basic_sse12_fp_binop_p_int<bits<8> opc, string OpcodeStr, + SizeItins itins, bit Is2Addr = 1> { defm PS : sse12_fp_packed_int<opc, OpcodeStr, VR128, !strconcat(OpcodeStr, "ps"), "sse", "_ps", f128mem, memopv4f32, - SSEPackedSingle, Is2Addr>, TB; + SSEPackedSingle, itins.s, Is2Addr>, + TB; defm PD : sse12_fp_packed_int<opc, OpcodeStr, VR128, !strconcat(OpcodeStr, "pd"), "sse2", "_pd", f128mem, memopv2f64, - SSEPackedDouble, Is2Addr>, TB, OpSize; + SSEPackedDouble, itins.d, Is2Addr>, + TB, OpSize; } -multiclass basic_sse12_fp_binop_p_y_int<bits<8> opc, string OpcodeStr> { +multiclass basic_sse12_fp_binop_p_y_int<bits<8> opc, string OpcodeStr, + SizeItins itins> { defm PSY : sse12_fp_packed_int<opc, OpcodeStr, VR256, !strconcat(OpcodeStr, "ps"), "avx", "_ps_256", f256mem, memopv8f32, - SSEPackedSingle, 0>, TB; + SSEPackedSingle, itins.s, 0>, TB; defm PDY : sse12_fp_packed_int<opc, OpcodeStr, VR256, !strconcat(OpcodeStr, "pd"), "avx", "_pd_256", f256mem, memopv4f64, - SSEPackedDouble, 0>, TB, OpSize; + SSEPackedDouble, itins.d, 0>, TB, OpSize; } // Binary Arithmetic instructions -defm VADD : basic_sse12_fp_binop_s<0x58, "add", fadd, 0>, - basic_sse12_fp_binop_s_int<0x58, "add", 0>, VEX_4V, VEX_LIG; -defm VADD : basic_sse12_fp_binop_p<0x58, "add", fadd, 0>, - basic_sse12_fp_binop_p_y<0x58, "add", fadd>, VEX_4V; -defm VMUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, 0>, - basic_sse12_fp_binop_s_int<0x59, "mul", 0>, VEX_4V, VEX_LIG; -defm VMUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, 0>, - basic_sse12_fp_binop_p_y<0x59, "mul", fmul>, VEX_4V; +defm VADD : basic_sse12_fp_binop_s<0x58, "add", fadd, SSE_ALU_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x58, "add", SSE_ALU_ITINS_S, 0>, + VEX_4V, VEX_LIG; +defm VADD : basic_sse12_fp_binop_p<0x58, "add", fadd, SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_y<0x58, "add", fadd, SSE_ALU_ITINS_P>, + VEX_4V; +defm VMUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, SSE_MUL_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x59, "mul", SSE_MUL_ITINS_S, 0>, + VEX_4V, VEX_LIG; +defm VMUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, SSE_MUL_ITINS_P, 0>, + basic_sse12_fp_binop_p_y<0x59, "mul", fmul, SSE_MUL_ITINS_P>, + VEX_4V; let isCommutable = 0 in { - defm VSUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, 0>, - basic_sse12_fp_binop_s_int<0x5C, "sub", 0>, VEX_4V, VEX_LIG; - defm VSUB : basic_sse12_fp_binop_p<0x5C, "sub", fsub, 0>, - basic_sse12_fp_binop_p_y<0x5C, "sub", fsub>, VEX_4V; - defm VDIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, 0>, - basic_sse12_fp_binop_s_int<0x5E, "div", 0>, VEX_4V, VEX_LIG; - defm VDIV : basic_sse12_fp_binop_p<0x5E, "div", fdiv, 0>, - basic_sse12_fp_binop_p_y<0x5E, "div", fdiv>, VEX_4V; - defm VMAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, 0>, - basic_sse12_fp_binop_s_int<0x5F, "max", 0>, VEX_4V, VEX_LIG; - defm VMAX : basic_sse12_fp_binop_p<0x5F, "max", X86fmax, 0>, - basic_sse12_fp_binop_p_int<0x5F, "max", 0>, - basic_sse12_fp_binop_p_y<0x5F, "max", X86fmax>, - basic_sse12_fp_binop_p_y_int<0x5F, "max">, VEX_4V; - defm VMIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, 0>, - basic_sse12_fp_binop_s_int<0x5D, "min", 0>, VEX_4V, VEX_LIG; - defm VMIN : basic_sse12_fp_binop_p<0x5D, "min", X86fmin, 0>, - basic_sse12_fp_binop_p_int<0x5D, "min", 0>, - basic_sse12_fp_binop_p_y_int<0x5D, "min">, - basic_sse12_fp_binop_p_y<0x5D, "min", X86fmin>, VEX_4V; + defm VSUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, SSE_ALU_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x5C, "sub", SSE_ALU_ITINS_S, 0>, + VEX_4V, VEX_LIG; + defm VSUB : basic_sse12_fp_binop_p<0x5C, "sub", fsub, SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_y<0x5C, "sub", fsub, SSE_ALU_ITINS_P>, VEX_4V; + defm VDIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, SSE_DIV_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x5E, "div", SSE_DIV_ITINS_S, 0>, + VEX_4V, VEX_LIG; + defm VDIV : basic_sse12_fp_binop_p<0x5E, "div", fdiv, SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_y<0x5E, "div", fdiv, SSE_DIV_ITINS_P>, + VEX_4V; + defm VMAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, SSE_ALU_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x5F, "max", SSE_ALU_ITINS_S, 0>, + VEX_4V, VEX_LIG; + defm VMAX : basic_sse12_fp_binop_p<0x5F, "max", X86fmax, SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_int<0x5F, "max", SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_y<0x5F, "max", X86fmax, SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_p_y_int<0x5F, "max", SSE_ALU_ITINS_P>, + VEX_4V; + defm VMIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, SSE_ALU_ITINS_S, 0>, + basic_sse12_fp_binop_s_int<0x5D, "min", SSE_ALU_ITINS_S, 0>, + VEX_4V, VEX_LIG; + defm VMIN : basic_sse12_fp_binop_p<0x5D, "min", X86fmin, SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_int<0x5D, "min", SSE_ALU_ITINS_P, 0>, + basic_sse12_fp_binop_p_y_int<0x5D, "min", SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_p_y<0x5D, "min", X86fmin, SSE_ALU_ITINS_P>, + VEX_4V; } let Constraints = "$src1 = $dst" in { - defm ADD : basic_sse12_fp_binop_s<0x58, "add", fadd>, - basic_sse12_fp_binop_p<0x58, "add", fadd>, - basic_sse12_fp_binop_s_int<0x58, "add">; - defm MUL : basic_sse12_fp_binop_s<0x59, "mul", fmul>, - basic_sse12_fp_binop_p<0x59, "mul", fmul>, - basic_sse12_fp_binop_s_int<0x59, "mul">; + defm ADD : basic_sse12_fp_binop_s<0x58, "add", fadd, SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p<0x58, "add", fadd, SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_s_int<0x58, "add", SSE_ALU_ITINS_S>; + defm MUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, SSE_MUL_ITINS_S>, + basic_sse12_fp_binop_p<0x59, "mul", fmul, SSE_MUL_ITINS_P>, + basic_sse12_fp_binop_s_int<0x59, "mul", SSE_MUL_ITINS_S>; let isCommutable = 0 in { - defm SUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub>, - basic_sse12_fp_binop_p<0x5C, "sub", fsub>, - basic_sse12_fp_binop_s_int<0x5C, "sub">; - defm DIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv>, - basic_sse12_fp_binop_p<0x5E, "div", fdiv>, - basic_sse12_fp_binop_s_int<0x5E, "div">; - defm MAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax>, - basic_sse12_fp_binop_p<0x5F, "max", X86fmax>, - basic_sse12_fp_binop_s_int<0x5F, "max">, - basic_sse12_fp_binop_p_int<0x5F, "max">; - defm MIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin>, - basic_sse12_fp_binop_p<0x5D, "min", X86fmin>, - basic_sse12_fp_binop_s_int<0x5D, "min">, - basic_sse12_fp_binop_p_int<0x5D, "min">; + defm SUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p<0x5C, "sub", fsub, SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_s_int<0x5C, "sub", SSE_ALU_ITINS_S>; + defm DIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, SSE_DIV_ITINS_S>, + basic_sse12_fp_binop_p<0x5E, "div", fdiv, SSE_DIV_ITINS_P>, + basic_sse12_fp_binop_s_int<0x5E, "div", SSE_DIV_ITINS_S>; + defm MAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p<0x5F, "max", X86fmax, SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_s_int<0x5F, "max", SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p_int<0x5F, "max", SSE_ALU_ITINS_P>; + defm MIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p<0x5D, "min", X86fmin, SSE_ALU_ITINS_P>, + basic_sse12_fp_binop_s_int<0x5D, "min", SSE_ALU_ITINS_S>, + basic_sse12_fp_binop_p_int<0x5D, "min", SSE_ALU_ITINS_P>; } } @@ -2607,9 +2976,25 @@ let Constraints = "$src1 = $dst" in { /// /// And, we have a special variant form for a full-vector intrinsic form. +def SSE_SQRTP : OpndItins< + IIC_SSE_SQRTP_RR, IIC_SSE_SQRTP_RM +>; + +def SSE_SQRTS : OpndItins< + IIC_SSE_SQRTS_RR, IIC_SSE_SQRTS_RM +>; + +def SSE_RCPP : OpndItins< + IIC_SSE_RCPP_RR, IIC_SSE_RCPP_RM +>; + +def SSE_RCPS : OpndItins< + IIC_SSE_RCPS_RR, IIC_SSE_RCPS_RM +>; + /// sse1_fp_unop_s - SSE1 unops in scalar form. multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr, - SDNode OpNode, Intrinsic F32Int> { + SDNode OpNode, Intrinsic F32Int, OpndItins itins> { def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src), !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"), [(set FR32:$dst, (OpNode FR32:$src))]>; @@ -2619,14 +3004,14 @@ multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr, // partial register update condition. def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src), !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"), - [(set FR32:$dst, (OpNode (load addr:$src)))]>, XS, + [(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS, Requires<[HasSSE1, OptForSize]>; def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (F32Int VR128:$src))]>; + [(set VR128:$dst, (F32Int VR128:$src))], itins.rr>; def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst), (ins ssmem:$src), !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (F32Int sse_load_f32:$src))]>; + [(set VR128:$dst, (F32Int sse_load_f32:$src))], itins.rm>; } /// sse1_fp_unop_s_avx - AVX SSE1 unops in scalar form. @@ -2645,64 +3030,72 @@ multiclass sse1_fp_unop_s_avx<bits<8> opc, string OpcodeStr> { } /// sse1_fp_unop_p - SSE1 unops in packed form. -multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode> { +multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode, + OpndItins itins> { def PSr : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))]>; + [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))], itins.rr>; def PSm : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))]>; + [(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))], itins.rm>; } /// sse1_fp_unop_p_y - AVX 256-bit SSE1 unops in packed form. -multiclass sse1_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode> { +multiclass sse1_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode, + OpndItins itins> { def PSYr : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (v8f32 (OpNode VR256:$src)))]>; + [(set VR256:$dst, (v8f32 (OpNode VR256:$src)))], + itins.rr>; def PSYm : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))]>; + [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))], + itins.rm>; } /// sse1_fp_unop_p_int - SSE1 intrinsics unops in packed forms. multiclass sse1_fp_unop_p_int<bits<8> opc, string OpcodeStr, - Intrinsic V4F32Int> { + Intrinsic V4F32Int, OpndItins itins> { def PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (V4F32Int VR128:$src))]>; + [(set VR128:$dst, (V4F32Int VR128:$src))], + itins.rr>; def PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))]>; + [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))], + itins.rm>; } /// sse1_fp_unop_p_y_int - AVX 256-bit intrinsics unops in packed forms. multiclass sse1_fp_unop_p_y_int<bits<8> opc, string OpcodeStr, - Intrinsic V4F32Int> { + Intrinsic V4F32Int, OpndItins itins> { def PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (V4F32Int VR256:$src))]>; + [(set VR256:$dst, (V4F32Int VR256:$src))], + itins.rr>; def PSYm_Int : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (V4F32Int (memopv8f32 addr:$src)))]>; + [(set VR256:$dst, (V4F32Int (memopv8f32 addr:$src)))], + itins.rm>; } /// sse2_fp_unop_s - SSE2 unops in scalar form. multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr, - SDNode OpNode, Intrinsic F64Int> { + SDNode OpNode, Intrinsic F64Int, OpndItins itins> { def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src), !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"), - [(set FR64:$dst, (OpNode FR64:$src))]>; + [(set FR64:$dst, (OpNode FR64:$src))], itins.rr>; // See the comments in sse1_fp_unop_s for why this is OptForSize. def SDm : I<opc, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src), !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"), - [(set FR64:$dst, (OpNode (load addr:$src)))]>, XD, + [(set FR64:$dst, (OpNode (load addr:$src)))], itins.rm>, XD, Requires<[HasSSE2, OptForSize]>; def SDr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (F64Int VR128:$src))]>; + [(set VR128:$dst, (F64Int VR128:$src))], itins.rr>; def SDm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst), (ins sdmem:$src), !strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (F64Int sse_load_f64:$src))]>; + [(set VR128:$dst, (F64Int sse_load_f64:$src))], itins.rm>; } /// sse2_fp_unop_s_avx - AVX SSE2 unops in scalar form. @@ -2724,45 +3117,52 @@ multiclass sse2_fp_unop_s_avx<bits<8> opc, string OpcodeStr> { /// sse2_fp_unop_p - SSE2 unops in vector forms. multiclass sse2_fp_unop_p<bits<8> opc, string OpcodeStr, - SDNode OpNode> { + SDNode OpNode, OpndItins itins> { def PDr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (v2f64 (OpNode VR128:$src)))]>; + [(set VR128:$dst, (v2f64 (OpNode VR128:$src)))], itins.rr>; def PDm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))]>; + [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))], itins.rm>; } /// sse2_fp_unop_p_y - AVX SSE2 256-bit unops in vector forms. -multiclass sse2_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode> { +multiclass sse2_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode, + OpndItins itins> { def PDYr : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (v4f64 (OpNode VR256:$src)))]>; + [(set VR256:$dst, (v4f64 (OpNode VR256:$src)))], + itins.rr>; def PDYm : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))]>; + [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))], + itins.rm>; } /// sse2_fp_unop_p_int - SSE2 intrinsic unops in vector forms. multiclass sse2_fp_unop_p_int<bits<8> opc, string OpcodeStr, - Intrinsic V2F64Int> { + Intrinsic V2F64Int, OpndItins itins> { def PDr_Int : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (V2F64Int VR128:$src))]>; + [(set VR128:$dst, (V2F64Int VR128:$src))], + itins.rr>; def PDm_Int : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (V2F64Int (memopv2f64 addr:$src)))]>; + [(set VR128:$dst, (V2F64Int (memopv2f64 addr:$src)))], + itins.rm>; } /// sse2_fp_unop_p_y_int - AVX 256-bit intrinsic unops in vector forms. multiclass sse2_fp_unop_p_y_int<bits<8> opc, string OpcodeStr, - Intrinsic V2F64Int> { + Intrinsic V2F64Int, OpndItins itins> { def PDYr_Int : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (V2F64Int VR256:$src))]>; + [(set VR256:$dst, (V2F64Int VR256:$src))], + itins.rr>; def PDYm_Int : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src), !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"), - [(set VR256:$dst, (V2F64Int (memopv4f64 addr:$src)))]>; + [(set VR256:$dst, (V2F64Int (memopv4f64 addr:$src)))], + itins.rm>; } let Predicates = [HasAVX] in { @@ -2770,29 +3170,37 @@ let Predicates = [HasAVX] in { defm VSQRT : sse1_fp_unop_s_avx<0x51, "vsqrt">, sse2_fp_unop_s_avx<0x51, "vsqrt">, VEX_4V, VEX_LIG; - defm VSQRT : sse1_fp_unop_p<0x51, "vsqrt", fsqrt>, - sse2_fp_unop_p<0x51, "vsqrt", fsqrt>, - sse1_fp_unop_p_y<0x51, "vsqrt", fsqrt>, - sse2_fp_unop_p_y<0x51, "vsqrt", fsqrt>, - sse1_fp_unop_p_int<0x51, "vsqrt", int_x86_sse_sqrt_ps>, - sse2_fp_unop_p_int<0x51, "vsqrt", int_x86_sse2_sqrt_pd>, - sse1_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_ps_256>, - sse2_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_pd_256>, + defm VSQRT : sse1_fp_unop_p<0x51, "vsqrt", fsqrt, SSE_SQRTP>, + sse2_fp_unop_p<0x51, "vsqrt", fsqrt, SSE_SQRTP>, + sse1_fp_unop_p_y<0x51, "vsqrt", fsqrt, SSE_SQRTP>, + sse2_fp_unop_p_y<0x51, "vsqrt", fsqrt, SSE_SQRTP>, + sse1_fp_unop_p_int<0x51, "vsqrt", int_x86_sse_sqrt_ps, + SSE_SQRTP>, + sse2_fp_unop_p_int<0x51, "vsqrt", int_x86_sse2_sqrt_pd, + SSE_SQRTP>, + sse1_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_ps_256, + SSE_SQRTP>, + sse2_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_pd_256, + SSE_SQRTP>, VEX; // Reciprocal approximations. Note that these typically require refinement // in order to obtain suitable precision. defm VRSQRT : sse1_fp_unop_s_avx<0x52, "vrsqrt">, VEX_4V, VEX_LIG; - defm VRSQRT : sse1_fp_unop_p<0x52, "vrsqrt", X86frsqrt>, - sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt>, - sse1_fp_unop_p_y_int<0x52, "vrsqrt", int_x86_avx_rsqrt_ps_256>, - sse1_fp_unop_p_int<0x52, "vrsqrt", int_x86_sse_rsqrt_ps>, VEX; + defm VRSQRT : sse1_fp_unop_p<0x52, "vrsqrt", X86frsqrt, SSE_SQRTP>, + sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt, SSE_SQRTP>, + sse1_fp_unop_p_y_int<0x52, "vrsqrt", int_x86_avx_rsqrt_ps_256, + SSE_SQRTP>, + sse1_fp_unop_p_int<0x52, "vrsqrt", int_x86_sse_rsqrt_ps, + SSE_SQRTP>, VEX; defm VRCP : sse1_fp_unop_s_avx<0x53, "vrcp">, VEX_4V, VEX_LIG; - defm VRCP : sse1_fp_unop_p<0x53, "vrcp", X86frcp>, - sse1_fp_unop_p_y<0x53, "vrcp", X86frcp>, - sse1_fp_unop_p_y_int<0x53, "vrcp", int_x86_avx_rcp_ps_256>, - sse1_fp_unop_p_int<0x53, "vrcp", int_x86_sse_rcp_ps>, VEX; + defm VRCP : sse1_fp_unop_p<0x53, "vrcp", X86frcp, SSE_RCPP>, + sse1_fp_unop_p_y<0x53, "vrcp", X86frcp, SSE_RCPP>, + sse1_fp_unop_p_y_int<0x53, "vrcp", int_x86_avx_rcp_ps_256, + SSE_RCPP>, + sse1_fp_unop_p_int<0x53, "vrcp", int_x86_sse_rcp_ps, + SSE_RCPP>, VEX; } let AddedComplexity = 1 in { @@ -2855,21 +3263,26 @@ let Predicates = [HasAVX], AddedComplexity = 1 in { } // Square root. -defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss>, - sse1_fp_unop_p<0x51, "sqrt", fsqrt>, - sse1_fp_unop_p_int<0x51, "sqrt", int_x86_sse_sqrt_ps>, - sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd>, - sse2_fp_unop_p<0x51, "sqrt", fsqrt>, - sse2_fp_unop_p_int<0x51, "sqrt", int_x86_sse2_sqrt_pd>; +defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss, + SSE_SQRTS>, + sse1_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTS>, + sse1_fp_unop_p_int<0x51, "sqrt", int_x86_sse_sqrt_ps, SSE_SQRTS>, + sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd, + SSE_SQRTS>, + sse2_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTS>, + sse2_fp_unop_p_int<0x51, "sqrt", int_x86_sse2_sqrt_pd, SSE_SQRTS>; // Reciprocal approximations. Note that these typically require refinement // in order to obtain suitable precision. -defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, int_x86_sse_rsqrt_ss>, - sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt>, - sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps>; -defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, int_x86_sse_rcp_ss>, - sse1_fp_unop_p<0x53, "rcp", X86frcp>, - sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps>; +defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, int_x86_sse_rsqrt_ss, + SSE_SQRTS>, + sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_SQRTS>, + sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps, + SSE_SQRTS>; +defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, int_x86_sse_rcp_ss, + SSE_RCPS>, + sse1_fp_unop_p<0x53, "rcp", X86frcp, SSE_RCPS>, + sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps, SSE_RCPS>; // There is no f64 version of the reciprocal approximation instructions. @@ -2882,19 +3295,22 @@ let AddedComplexity = 400 in { // Prefer non-temporal versions (ins f128mem:$dst, VR128:$src), "movntps\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v4f32 VR128:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; def VMOVNTPDmr : VPDI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movntpd\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v2f64 VR128:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; let ExeDomain = SSEPackedInt in def VMOVNTDQmr : VPDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movntdq\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v2i64 VR128:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst), (VMOVNTDQmr addr:$dst, VR128:$src)>, Requires<[HasAVX]>; @@ -2903,18 +3319,21 @@ let AddedComplexity = 400 in { // Prefer non-temporal versions (ins f256mem:$dst, VR256:$src), "movntps\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v8f32 VR256:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; def VMOVNTPDYmr : VPDI<0x2B, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movntpd\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v4f64 VR256:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; let ExeDomain = SSEPackedInt in def VMOVNTDQYmr : VPDI<0xE7, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src), "movntdq\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v4i64 VR256:$src), - addr:$dst)]>, VEX; + addr:$dst)], + IIC_SSE_MOVNT>, VEX; } def : Pat<(int_x86_avx_movnt_dq_256 addr:$dst, VR256:$src), @@ -2927,15 +3346,18 @@ def : Pat<(int_x86_avx_movnt_ps_256 addr:$dst, VR256:$src), let AddedComplexity = 400 in { // Prefer non-temporal versions def MOVNTPSmr : PSI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movntps\t{$src, $dst|$dst, $src}", - [(alignednontemporalstore (v4f32 VR128:$src), addr:$dst)]>; + [(alignednontemporalstore (v4f32 VR128:$src), addr:$dst)], + IIC_SSE_MOVNT>; def MOVNTPDmr : PDI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movntpd\t{$src, $dst|$dst, $src}", - [(alignednontemporalstore(v2f64 VR128:$src), addr:$dst)]>; + [(alignednontemporalstore(v2f64 VR128:$src), addr:$dst)], + IIC_SSE_MOVNT>; let ExeDomain = SSEPackedInt in def MOVNTDQmr : PDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src), "movntdq\t{$src, $dst|$dst, $src}", - [(alignednontemporalstore (v2i64 VR128:$src), addr:$dst)]>; + [(alignednontemporalstore (v2i64 VR128:$src), addr:$dst)], + IIC_SSE_MOVNT>; def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst), (MOVNTDQmr addr:$dst, VR128:$src)>, Requires<[HasSSE2]>; @@ -2943,11 +3365,13 @@ def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst), // There is no AVX form for instructions below this point def MOVNTImr : I<0xC3, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src), "movnti{l}\t{$src, $dst|$dst, $src}", - [(nontemporalstore (i32 GR32:$src), addr:$dst)]>, + [(nontemporalstore (i32 GR32:$src), addr:$dst)], + IIC_SSE_MOVNT>, TB, Requires<[HasSSE2]>; def MOVNTI_64mr : RI<0xC3, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), "movnti{q}\t{$src, $dst|$dst, $src}", - [(nontemporalstore (i64 GR64:$src), addr:$dst)]>, + [(nontemporalstore (i64 GR64:$src), addr:$dst)], + IIC_SSE_MOVNT>, TB, Requires<[HasSSE2]>; } @@ -2958,31 +3382,38 @@ def MOVNTI_64mr : RI<0xC3, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src), // Prefetch intrinsic. let Predicates = [HasSSE1] in { def PREFETCHT0 : I<0x18, MRM1m, (outs), (ins i8mem:$src), - "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))]>, TB; + "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))], + IIC_SSE_PREFETCH>, TB; def PREFETCHT1 : I<0x18, MRM2m, (outs), (ins i8mem:$src), - "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))]>, TB; + "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))], + IIC_SSE_PREFETCH>, TB; def PREFETCHT2 : I<0x18, MRM3m, (outs), (ins i8mem:$src), - "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))]>, TB; + "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))], + IIC_SSE_PREFETCH>, TB; def PREFETCHNTA : I<0x18, MRM0m, (outs), (ins i8mem:$src), - "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))]>, TB; + "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))], + IIC_SSE_PREFETCH>, TB; } // Flush cache def CLFLUSH : I<0xAE, MRM7m, (outs), (ins i8mem:$src), - "clflush\t$src", [(int_x86_sse2_clflush addr:$src)]>, - TB, Requires<[HasSSE2]>; + "clflush\t$src", [(int_x86_sse2_clflush addr:$src)], + IIC_SSE_PREFETCH>, TB, Requires<[HasSSE2]>; // Pause. This "instruction" is encoded as "rep; nop", so even though it // was introduced with SSE2, it's backward compatible. -def PAUSE : I<0x90, RawFrm, (outs), (ins), "pause", []>, REP; +def PAUSE : I<0x90, RawFrm, (outs), (ins), "pause", [], IIC_SSE_PAUSE>, REP; // Load, store, and memory fence def SFENCE : I<0xAE, MRM_F8, (outs), (ins), - "sfence", [(int_x86_sse_sfence)]>, TB, Requires<[HasSSE1]>; + "sfence", [(int_x86_sse_sfence)], IIC_SSE_SFENCE>, + TB, Requires<[HasSSE1]>; def LFENCE : I<0xAE, MRM_E8, (outs), (ins), - "lfence", [(int_x86_sse2_lfence)]>, TB, Requires<[HasSSE2]>; + "lfence", [(int_x86_sse2_lfence)], IIC_SSE_LFENCE>, + TB, Requires<[HasSSE2]>; def MFENCE : I<0xAE, MRM_F0, (outs), (ins), - "mfence", [(int_x86_sse2_mfence)]>, TB, Requires<[HasSSE2]>; + "mfence", [(int_x86_sse2_mfence)], IIC_SSE_MFENCE>, + TB, Requires<[HasSSE2]>; def : Pat<(X86SFence), (SFENCE)>; def : Pat<(X86LFence), (LFENCE)>; @@ -2993,14 +3424,18 @@ def : Pat<(X86MFence), (MFENCE)>; //===----------------------------------------------------------------------===// def VLDMXCSR : VPSI<0xAE, MRM2m, (outs), (ins i32mem:$src), - "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)]>, VEX; + "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)], + IIC_SSE_LDMXCSR>, VEX; def VSTMXCSR : VPSI<0xAE, MRM3m, (outs), (ins i32mem:$dst), - "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)]>, VEX; + "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)], + IIC_SSE_STMXCSR>, VEX; def LDMXCSR : PSI<0xAE, MRM2m, (outs), (ins i32mem:$src), - "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)]>; + "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)], + IIC_SSE_LDMXCSR>; def STMXCSR : PSI<0xAE, MRM3m, (outs), (ins i32mem:$dst), - "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)]>; + "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)], + IIC_SSE_STMXCSR>; //===---------------------------------------------------------------------===// // SSE2 - Move Aligned/Unaligned Packed Integer Instructions @@ -3010,102 +3445,129 @@ let ExeDomain = SSEPackedInt in { // SSE integer instructions let neverHasSideEffects = 1 in { def VMOVDQArr : VPDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>, + VEX; def VMOVDQAYrr : VPDI<0x6F, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>, + VEX; } def VMOVDQUrr : VSSI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "movdqu\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqu\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVU_P_RR>, + VEX; def VMOVDQUYrr : VSSI<0x6F, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src), - "movdqu\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqu\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVU_P_RR>, + VEX; // For Disassembler let isCodeGenOnly = 1 in { def VMOVDQArr_REV : VPDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, + VEX; def VMOVDQAYrr_REV : VPDI<0x7F, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>, + VEX; def VMOVDQUrr_REV : VSSI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movdqu\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqu\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, + VEX; def VMOVDQUYrr_REV : VSSI<0x7F, MRMDestReg, (outs VR256:$dst), (ins VR256:$src), - "movdqu\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqu\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVU_P_RR>, + VEX; } let canFoldAsLoad = 1, mayLoad = 1 in { def VMOVDQArm : VPDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>, + VEX; def VMOVDQAYrm : VPDI<0x6F, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>, + VEX; let Predicates = [HasAVX] in { def VMOVDQUrm : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), - "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX; + "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_RM>, + XS, VEX; def VMOVDQUYrm : I<0x6F, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src), - "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX; + "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_RM>, + XS, VEX; } } let mayStore = 1 in { def VMOVDQAmr : VPDI<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>, + VEX; def VMOVDQAYmr : VPDI<0x7F, MRMDestMem, (outs), (ins i256mem:$dst, VR256:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>, VEX; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>, + VEX; let Predicates = [HasAVX] in { def VMOVDQUmr : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), - "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX; + "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_MR>, + XS, VEX; def VMOVDQUYmr : I<0x7F, MRMDestMem, (outs), (ins i256mem:$dst, VR256:$src), - "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX; + "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_MR>, + XS, VEX; } } let neverHasSideEffects = 1 in def MOVDQArr : PDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>; + "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>; def MOVDQUrr : I<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "movdqu\t{$src, $dst|$dst, $src}", - []>, XS, Requires<[HasSSE2]>; + [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>; // For Disassembler let isCodeGenOnly = 1 in { def MOVDQArr_REV : PDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), - "movdqa\t{$src, $dst|$dst, $src}", []>; + "movdqa\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVA_P_RR>; def MOVDQUrr_REV : I<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src), "movdqu\t{$src, $dst|$dst, $src}", - []>, XS, Requires<[HasSSE2]>; + [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>; } let canFoldAsLoad = 1, mayLoad = 1 in { def MOVDQArm : PDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "movdqa\t{$src, $dst|$dst, $src}", - [/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/]>; + [/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/], + IIC_SSE_MOVA_P_RM>; def MOVDQUrm : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "movdqu\t{$src, $dst|$dst, $src}", - [/*(set VR128:$dst, (loadv2i64 addr:$src))*/]>, + [/*(set VR128:$dst, (loadv2i64 addr:$src))*/], + IIC_SSE_MOVU_P_RM>, XS, Requires<[HasSSE2]>; } let mayStore = 1 in { def MOVDQAmr : PDI<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), "movdqa\t{$src, $dst|$dst, $src}", - [/*(alignedstore (v2i64 VR128:$src), addr:$dst)*/]>; + [/*(alignedstore (v2i64 VR128:$src), addr:$dst)*/], + IIC_SSE_MOVA_P_MR>; def MOVDQUmr : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), "movdqu\t{$src, $dst|$dst, $src}", - [/*(store (v2i64 VR128:$src), addr:$dst)*/]>, + [/*(store (v2i64 VR128:$src), addr:$dst)*/], + IIC_SSE_MOVU_P_MR>, XS, Requires<[HasSSE2]>; } // Intrinsic forms of MOVDQU load and store def VMOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), "vmovdqu\t{$src, $dst|$dst, $src}", - [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)]>, + [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)], + IIC_SSE_MOVU_P_MR>, XS, VEX, Requires<[HasAVX]>; def MOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src), "movdqu\t{$src, $dst|$dst, $src}", - [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)]>, + [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)], + IIC_SSE_MOVU_P_MR>, XS, Requires<[HasSSE2]>; } // ExeDomain = SSEPackedInt @@ -3119,11 +3581,17 @@ let Predicates = [HasAVX] in { // SSE2 - Packed Integer Arithmetic Instructions //===---------------------------------------------------------------------===// +def SSE_PMADD : OpndItins< + IIC_SSE_PMADD, IIC_SSE_PMADD +>; + let ExeDomain = SSEPackedInt in { // SSE integer instructions multiclass PDI_binop_rm_int<bits<8> opc, string OpcodeStr, Intrinsic IntId, RegisterClass RC, PatFrag memop_frag, - X86MemOperand x86memop, bit IsCommutable = 0, + X86MemOperand x86memop, + OpndItins itins, + bit IsCommutable = 0, bit Is2Addr = 1> { let isCommutable = IsCommutable in def rr : PDI<opc, MRMSrcReg, (outs RC:$dst), @@ -3131,19 +3599,21 @@ multiclass PDI_binop_rm_int<bits<8> opc, string OpcodeStr, Intrinsic IntId, !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (IntId RC:$src1, RC:$src2))]>; + [(set RC:$dst, (IntId RC:$src1, RC:$src2))], itins.rr>; def rm : PDI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (IntId RC:$src1, (bitconvert (memop_frag addr:$src2))))]>; + [(set RC:$dst, (IntId RC:$src1, (bitconvert (memop_frag addr:$src2))))], + itins.rm>; } multiclass PDI_binop_rmi<bits<8> opc, bits<8> opc2, Format ImmForm, string OpcodeStr, SDNode OpNode, SDNode OpNode2, RegisterClass RC, ValueType DstVT, ValueType SrcVT, PatFrag bc_frag, + ShiftOpndItins itins, bit Is2Addr = 1> { // src2 is always 128-bit def rr : PDI<opc, MRMSrcReg, (outs RC:$dst), @@ -3151,26 +3621,28 @@ multiclass PDI_binop_rmi<bits<8> opc, bits<8> opc2, Format ImmForm, !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (DstVT (OpNode RC:$src1, (SrcVT VR128:$src2))))]>; + [(set RC:$dst, (DstVT (OpNode RC:$src1, (SrcVT VR128:$src2))))], + itins.rr>; def rm : PDI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, i128mem:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (DstVT (OpNode RC:$src1, - (bc_frag (memopv2i64 addr:$src2)))))]>; + (bc_frag (memopv2i64 addr:$src2)))))], itins.rm>; def ri : PDIi8<opc2, ImmForm, (outs RC:$dst), (ins RC:$src1, i32i8imm:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))]>; + [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))], itins.ri>; } /// PDI_binop_rm - Simple SSE2 binary operator with different src and dst types multiclass PDI_binop_rm2<bits<8> opc, string OpcodeStr, SDNode OpNode, ValueType DstVT, ValueType SrcVT, RegisterClass RC, PatFrag memop_frag, X86MemOperand x86memop, + OpndItins itins, bit IsCommutable = 0, bit Is2Addr = 1> { let isCommutable = IsCommutable in def rr : PDI<opc, MRMSrcReg, (outs RC:$dst), @@ -3193,185 +3665,242 @@ multiclass PDI_binop_rm2<bits<8> opc, string OpcodeStr, SDNode OpNode, let Predicates = [HasAVX] in { defm VPADDB : PDI_binop_rm<0xFC, "vpaddb", add, v16i8, VR128, memopv2i64, - i128mem, 1, 0 /*3addr*/>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 1, 0 /*3addr*/>, + VEX_4V; defm VPADDW : PDI_binop_rm<0xFD, "vpaddw", add, v8i16, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDD : PDI_binop_rm<0xFE, "vpaddd", add, v4i32, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDQ : PDI_binop_rm<0xD4, "vpaddq", add, v2i64, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V; defm VPMULLW : PDI_binop_rm<0xD5, "vpmullw", mul, v8i16, VR128, memopv2i64, - i128mem, 1, 0>, VEX_4V; + i128mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPSUBB : PDI_binop_rm<0xF8, "vpsubb", sub, v16i8, VR128, memopv2i64, - i128mem, 0, 0>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBW : PDI_binop_rm<0xF9, "vpsubw", sub, v8i16, VR128, memopv2i64, - i128mem, 0, 0>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBD : PDI_binop_rm<0xFA, "vpsubd", sub, v4i32, VR128, memopv2i64, - i128mem, 0, 0>, VEX_4V; + i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBQ : PDI_binop_rm<0xFB, "vpsubq", sub, v2i64, VR128, memopv2i64, - i128mem, 0, 0>, VEX_4V; + i128mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V; defm VPMULUDQ : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v2i64, v4i32, VR128, - memopv2i64, i128mem, 1, 0>, VEX_4V; + memopv2i64, i128mem, SSE_INTMUL_ITINS_P, 1, 0>, + VEX_4V; // Intrinsic forms defm VPSUBSB : PDI_binop_rm_int<0xE8, "vpsubsb" , int_x86_sse2_psubs_b, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBSW : PDI_binop_rm_int<0xE9, "vpsubsw" , int_x86_sse2_psubs_w, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBUSB : PDI_binop_rm_int<0xD8, "vpsubusb", int_x86_sse2_psubus_b, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBUSW : PDI_binop_rm_int<0xD9, "vpsubusw", int_x86_sse2_psubus_w, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPADDSB : PDI_binop_rm_int<0xEC, "vpaddsb" , int_x86_sse2_padds_b, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDSW : PDI_binop_rm_int<0xED, "vpaddsw" , int_x86_sse2_padds_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDUSB : PDI_binop_rm_int<0xDC, "vpaddusb", int_x86_sse2_paddus_b, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDUSW : PDI_binop_rm_int<0xDD, "vpaddusw", int_x86_sse2_paddus_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMULHUW : PDI_binop_rm_int<0xE4, "vpmulhuw", int_x86_sse2_pmulhu_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPMULHW : PDI_binop_rm_int<0xE5, "vpmulhw" , int_x86_sse2_pmulh_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPMADDWD : PDI_binop_rm_int<0xF5, "vpmaddwd", int_x86_sse2_pmadd_wd, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_PMADD, 1, 0>, VEX_4V; defm VPAVGB : PDI_binop_rm_int<0xE0, "vpavgb", int_x86_sse2_pavg_b, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPAVGW : PDI_binop_rm_int<0xE3, "vpavgw", int_x86_sse2_pavg_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMINUB : PDI_binop_rm_int<0xDA, "vpminub", int_x86_sse2_pminu_b, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMINSW : PDI_binop_rm_int<0xEA, "vpminsw", int_x86_sse2_pmins_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMAXUB : PDI_binop_rm_int<0xDE, "vpmaxub", int_x86_sse2_pmaxu_b, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMAXSW : PDI_binop_rm_int<0xEE, "vpmaxsw", int_x86_sse2_pmaxs_w, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPSADBW : PDI_binop_rm_int<0xF6, "vpsadbw", int_x86_sse2_psad_bw, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; } let Predicates = [HasAVX2] in { defm VPADDBY : PDI_binop_rm<0xFC, "vpaddb", add, v32i8, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDWY : PDI_binop_rm<0xFD, "vpaddw", add, v16i16, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDDY : PDI_binop_rm<0xFE, "vpaddd", add, v8i32, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDQY : PDI_binop_rm<0xD4, "vpaddq", add, v4i64, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V; defm VPMULLWY : PDI_binop_rm<0xD5, "vpmullw", mul, v16i16, VR256, memopv4i64, - i256mem, 1, 0>, VEX_4V; + i256mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPSUBBY : PDI_binop_rm<0xF8, "vpsubb", sub, v32i8, VR256, memopv4i64, - i256mem, 0, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBWY : PDI_binop_rm<0xF9, "vpsubw", sub, v16i16,VR256, memopv4i64, - i256mem, 0, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBDY : PDI_binop_rm<0xFA, "vpsubd", sub, v8i32, VR256, memopv4i64, - i256mem, 0, 0>, VEX_4V; + i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBQY : PDI_binop_rm<0xFB, "vpsubq", sub, v4i64, VR256, memopv4i64, - i256mem, 0, 0>, VEX_4V; + i256mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V; defm VPMULUDQY : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v4i64, v8i32, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; // Intrinsic forms defm VPSUBSBY : PDI_binop_rm_int<0xE8, "vpsubsb" , int_x86_avx2_psubs_b, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBSWY : PDI_binop_rm_int<0xE9, "vpsubsw" , int_x86_avx2_psubs_w, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBUSBY : PDI_binop_rm_int<0xD8, "vpsubusb", int_x86_avx2_psubus_b, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPSUBUSWY : PDI_binop_rm_int<0xD9, "vpsubusw", int_x86_avx2_psubus_w, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPADDSBY : PDI_binop_rm_int<0xEC, "vpaddsb" , int_x86_avx2_padds_b, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDSWY : PDI_binop_rm_int<0xED, "vpaddsw" , int_x86_avx2_padds_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDUSBY : PDI_binop_rm_int<0xDC, "vpaddusb", int_x86_avx2_paddus_b, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPADDUSWY : PDI_binop_rm_int<0xDD, "vpaddusw", int_x86_avx2_paddus_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMULHUWY : PDI_binop_rm_int<0xE4, "vpmulhuw", int_x86_avx2_pmulhu_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPMULHWY : PDI_binop_rm_int<0xE5, "vpmulhw" , int_x86_avx2_pmulh_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V; defm VPMADDWDY : PDI_binop_rm_int<0xF5, "vpmaddwd", int_x86_avx2_pmadd_wd, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_PMADD, 1, 0>, VEX_4V; defm VPAVGBY : PDI_binop_rm_int<0xE0, "vpavgb", int_x86_avx2_pavg_b, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPAVGWY : PDI_binop_rm_int<0xE3, "vpavgw", int_x86_avx2_pavg_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMINUBY : PDI_binop_rm_int<0xDA, "vpminub", int_x86_avx2_pminu_b, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMINSWY : PDI_binop_rm_int<0xEA, "vpminsw", int_x86_avx2_pmins_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMAXUBY : PDI_binop_rm_int<0xDE, "vpmaxub", int_x86_avx2_pmaxu_b, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPMAXSWY : PDI_binop_rm_int<0xEE, "vpmaxsw", int_x86_avx2_pmaxs_w, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPSADBWY : PDI_binop_rm_int<0xF6, "vpsadbw", int_x86_avx2_psad_bw, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; } let Constraints = "$src1 = $dst" in { defm PADDB : PDI_binop_rm<0xFC, "paddb", add, v16i8, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_INTALU_ITINS_P, 1>; defm PADDW : PDI_binop_rm<0xFD, "paddw", add, v8i16, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_INTALU_ITINS_P, 1>; defm PADDD : PDI_binop_rm<0xFE, "paddd", add, v4i32, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_INTALU_ITINS_P, 1>; defm PADDQ : PDI_binop_rm<0xD4, "paddq", add, v2i64, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_INTALUQ_ITINS_P, 1>; defm PMULLW : PDI_binop_rm<0xD5, "pmullw", mul, v8i16, VR128, memopv2i64, - i128mem, 1>; + i128mem, SSE_INTMUL_ITINS_P, 1>; defm PSUBB : PDI_binop_rm<0xF8, "psubb", sub, v16i8, VR128, memopv2i64, - i128mem>; + i128mem, SSE_INTALU_ITINS_P>; defm PSUBW : PDI_binop_rm<0xF9, "psubw", sub, v8i16, VR128, memopv2i64, - i128mem>; + i128mem, SSE_INTALU_ITINS_P>; defm PSUBD : PDI_binop_rm<0xFA, "psubd", sub, v4i32, VR128, memopv2i64, - i128mem>; + i128mem, SSE_INTALU_ITINS_P>; defm PSUBQ : PDI_binop_rm<0xFB, "psubq", sub, v2i64, VR128, memopv2i64, - i128mem>; + i128mem, SSE_INTALUQ_ITINS_P>; defm PMULUDQ : PDI_binop_rm2<0xF4, "pmuludq", X86pmuludq, v2i64, v4i32, VR128, - memopv2i64, i128mem, 1>; + memopv2i64, i128mem, SSE_INTMUL_ITINS_P, 1>; // Intrinsic forms defm PSUBSB : PDI_binop_rm_int<0xE8, "psubsb" , int_x86_sse2_psubs_b, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PSUBSW : PDI_binop_rm_int<0xE9, "psubsw" , int_x86_sse2_psubs_w, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PSUBUSB : PDI_binop_rm_int<0xD8, "psubusb", int_x86_sse2_psubus_b, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PSUBUSW : PDI_binop_rm_int<0xD9, "psubusw", int_x86_sse2_psubus_w, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PADDSB : PDI_binop_rm_int<0xEC, "paddsb" , int_x86_sse2_padds_b, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PADDSW : PDI_binop_rm_int<0xED, "paddsw" , int_x86_sse2_padds_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PADDUSB : PDI_binop_rm_int<0xDC, "paddusb", int_x86_sse2_paddus_b, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PADDUSW : PDI_binop_rm_int<0xDD, "paddusw", int_x86_sse2_paddus_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PMULHUW : PDI_binop_rm_int<0xE4, "pmulhuw", int_x86_sse2_pmulhu_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTMUL_ITINS_P, 1>; defm PMULHW : PDI_binop_rm_int<0xE5, "pmulhw" , int_x86_sse2_pmulh_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTMUL_ITINS_P, 1>; defm PMADDWD : PDI_binop_rm_int<0xF5, "pmaddwd", int_x86_sse2_pmadd_wd, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_PMADD, 1>; defm PAVGB : PDI_binop_rm_int<0xE0, "pavgb", int_x86_sse2_pavg_b, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PAVGW : PDI_binop_rm_int<0xE3, "pavgw", int_x86_sse2_pavg_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PMINUB : PDI_binop_rm_int<0xDA, "pminub", int_x86_sse2_pminu_b, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PMINSW : PDI_binop_rm_int<0xEA, "pminsw", int_x86_sse2_pmins_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PMAXUB : PDI_binop_rm_int<0xDE, "pmaxub", int_x86_sse2_pmaxu_b, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PMAXSW : PDI_binop_rm_int<0xEE, "pmaxsw", int_x86_sse2_pmaxs_w, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PSADBW : PDI_binop_rm_int<0xF6, "psadbw", int_x86_sse2_psad_bw, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; } // Constraints = "$src1 = $dst" @@ -3381,23 +3910,31 @@ defm PSADBW : PDI_binop_rm_int<0xF6, "psadbw", int_x86_sse2_psad_bw, let Predicates = [HasAVX] in { defm VPSLLW : PDI_binop_rmi<0xF1, 0x71, MRM6r, "vpsllw", X86vshl, X86vshli, - VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSLLD : PDI_binop_rmi<0xF2, 0x72, MRM6r, "vpslld", X86vshl, X86vshli, - VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSLLQ : PDI_binop_rmi<0xF3, 0x73, MRM6r, "vpsllq", X86vshl, X86vshli, - VR128, v2i64, v2i64, bc_v2i64, 0>, VEX_4V; + VR128, v2i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLW : PDI_binop_rmi<0xD1, 0x71, MRM2r, "vpsrlw", X86vsrl, X86vsrli, - VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLD : PDI_binop_rmi<0xD2, 0x72, MRM2r, "vpsrld", X86vsrl, X86vsrli, - VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLQ : PDI_binop_rmi<0xD3, 0x73, MRM2r, "vpsrlq", X86vsrl, X86vsrli, - VR128, v2i64, v2i64, bc_v2i64, 0>, VEX_4V; + VR128, v2i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRAW : PDI_binop_rmi<0xE1, 0x71, MRM4r, "vpsraw", X86vsra, X86vsrai, - VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRAD : PDI_binop_rmi<0xE2, 0x72, MRM4r, "vpsrad", X86vsra, X86vsrai, - VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; let ExeDomain = SSEPackedInt in { // 128-bit logical shifts. @@ -3419,23 +3956,31 @@ let ExeDomain = SSEPackedInt in { let Predicates = [HasAVX2] in { defm VPSLLWY : PDI_binop_rmi<0xF1, 0x71, MRM6r, "vpsllw", X86vshl, X86vshli, - VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR256, v16i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSLLDY : PDI_binop_rmi<0xF2, 0x72, MRM6r, "vpslld", X86vshl, X86vshli, - VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR256, v8i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSLLQY : PDI_binop_rmi<0xF3, 0x73, MRM6r, "vpsllq", X86vshl, X86vshli, - VR256, v4i64, v2i64, bc_v2i64, 0>, VEX_4V; + VR256, v4i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLWY : PDI_binop_rmi<0xD1, 0x71, MRM2r, "vpsrlw", X86vsrl, X86vsrli, - VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR256, v16i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLDY : PDI_binop_rmi<0xD2, 0x72, MRM2r, "vpsrld", X86vsrl, X86vsrli, - VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR256, v8i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRLQY : PDI_binop_rmi<0xD3, 0x73, MRM2r, "vpsrlq", X86vsrl, X86vsrli, - VR256, v4i64, v2i64, bc_v2i64, 0>, VEX_4V; + VR256, v4i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRAWY : PDI_binop_rmi<0xE1, 0x71, MRM4r, "vpsraw", X86vsra, X86vsrai, - VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V; + VR256, v16i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; defm VPSRADY : PDI_binop_rmi<0xE2, 0x72, MRM4r, "vpsrad", X86vsra, X86vsrai, - VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V; + VR256, v8i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P, 0>, VEX_4V; let ExeDomain = SSEPackedInt in { // 256-bit logical shifts. @@ -3457,23 +4002,31 @@ let ExeDomain = SSEPackedInt in { let Constraints = "$src1 = $dst" in { defm PSLLW : PDI_binop_rmi<0xF1, 0x71, MRM6r, "psllw", X86vshl, X86vshli, - VR128, v8i16, v8i16, bc_v8i16>; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P>; defm PSLLD : PDI_binop_rmi<0xF2, 0x72, MRM6r, "pslld", X86vshl, X86vshli, - VR128, v4i32, v4i32, bc_v4i32>; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P>; defm PSLLQ : PDI_binop_rmi<0xF3, 0x73, MRM6r, "psllq", X86vshl, X86vshli, - VR128, v2i64, v2i64, bc_v2i64>; + VR128, v2i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P>; defm PSRLW : PDI_binop_rmi<0xD1, 0x71, MRM2r, "psrlw", X86vsrl, X86vsrli, - VR128, v8i16, v8i16, bc_v8i16>; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P>; defm PSRLD : PDI_binop_rmi<0xD2, 0x72, MRM2r, "psrld", X86vsrl, X86vsrli, - VR128, v4i32, v4i32, bc_v4i32>; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P>; defm PSRLQ : PDI_binop_rmi<0xD3, 0x73, MRM2r, "psrlq", X86vsrl, X86vsrli, - VR128, v2i64, v2i64, bc_v2i64>; + VR128, v2i64, v2i64, bc_v2i64, + SSE_INTSHIFT_ITINS_P>; defm PSRAW : PDI_binop_rmi<0xE1, 0x71, MRM4r, "psraw", X86vsra, X86vsrai, - VR128, v8i16, v8i16, bc_v8i16>; + VR128, v8i16, v8i16, bc_v8i16, + SSE_INTSHIFT_ITINS_P>; defm PSRAD : PDI_binop_rmi<0xE2, 0x72, MRM4r, "psrad", X86vsra, X86vsrai, - VR128, v4i32, v4i32, bc_v4i32>; + VR128, v4i32, v4i32, bc_v4i32, + SSE_INTSHIFT_ITINS_P>; let ExeDomain = SSEPackedInt in { // 128-bit logical shifts. @@ -3534,47 +4087,65 @@ let Predicates = [HasSSE2] in { let Predicates = [HasAVX] in { defm VPCMPEQB : PDI_binop_rm<0x74, "vpcmpeqb", X86pcmpeq, v16i8, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPEQW : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v8i16, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPEQD : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v4i32, - VR128, memopv2i64, i128mem, 1, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPGTB : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v16i8, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPCMPGTW : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v8i16, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPCMPGTD : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v4i32, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; } let Predicates = [HasAVX2] in { defm VPCMPEQBY : PDI_binop_rm<0x74, "vpcmpeqb", X86pcmpeq, v32i8, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPEQWY : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v16i16, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPEQDY : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v8i32, - VR256, memopv4i64, i256mem, 1, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 1, 0>, VEX_4V; defm VPCMPGTBY : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v32i8, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPCMPGTWY : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v16i16, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPCMPGTDY : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v8i32, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; } let Constraints = "$src1 = $dst" in { defm PCMPEQB : PDI_binop_rm<0x74, "pcmpeqb", X86pcmpeq, v16i8, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PCMPEQW : PDI_binop_rm<0x75, "pcmpeqw", X86pcmpeq, v8i16, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PCMPEQD : PDI_binop_rm<0x76, "pcmpeqd", X86pcmpeq, v4i32, - VR128, memopv2i64, i128mem, 1>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 1>; defm PCMPGTB : PDI_binop_rm<0x64, "pcmpgtb", X86pcmpgt, v16i8, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PCMPGTW : PDI_binop_rm<0x65, "pcmpgtw", X86pcmpgt, v8i16, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PCMPGTD : PDI_binop_rm<0x66, "pcmpgtd", X86pcmpgt, v4i32, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; } // Constraints = "$src1 = $dst" //===---------------------------------------------------------------------===// @@ -3583,29 +4154,38 @@ let Constraints = "$src1 = $dst" in { let Predicates = [HasAVX] in { defm VPACKSSWB : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_sse2_packsswb_128, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPACKSSDW : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_sse2_packssdw_128, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPACKUSWB : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_sse2_packuswb_128, - VR128, memopv2i64, i128mem, 0, 0>, VEX_4V; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; } let Predicates = [HasAVX2] in { defm VPACKSSWBY : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_avx2_packsswb, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPACKSSDWY : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_avx2_packssdw, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; defm VPACKUSWBY : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_avx2_packuswb, - VR256, memopv4i64, i256mem, 0, 0>, VEX_4V; + VR256, memopv4i64, i256mem, + SSE_INTALU_ITINS_P, 0, 0>, VEX_4V; } let Constraints = "$src1 = $dst" in { defm PACKSSWB : PDI_binop_rm_int<0x63, "packsswb", int_x86_sse2_packsswb_128, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PACKSSDW : PDI_binop_rm_int<0x6B, "packssdw", int_x86_sse2_packssdw_128, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; defm PACKUSWB : PDI_binop_rm_int<0x67, "packuswb", int_x86_sse2_packuswb_128, - VR128, memopv2i64, i128mem>; + VR128, memopv2i64, i128mem, + SSE_INTALU_ITINS_P>; } // Constraints = "$src1 = $dst" //===---------------------------------------------------------------------===// @@ -3618,14 +4198,16 @@ def ri : Ii8<0x70, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, i8imm:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), - [(set VR128:$dst, (vt (OpNode VR128:$src1, (i8 imm:$src2))))]>; + [(set VR128:$dst, (vt (OpNode VR128:$src1, (i8 imm:$src2))))], + IIC_SSE_PSHUF>; def mi : Ii8<0x70, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set VR128:$dst, (vt (OpNode (bitconvert (memopv2i64 addr:$src1)), - (i8 imm:$src2))))]>; + (i8 imm:$src2))))], + IIC_SSE_PSHUF>; } multiclass sse2_pshuffle_y<string OpcodeStr, ValueType vt, SDNode OpNode> { @@ -3694,7 +4276,8 @@ multiclass sse2_unpack<bits<8> opc, string OpcodeStr, ValueType vt, !if(Is2Addr, !strconcat(OpcodeStr,"\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr,"\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set VR128:$dst, (vt (OpNode VR128:$src1, VR128:$src2)))]>; + [(set VR128:$dst, (vt (OpNode VR128:$src1, VR128:$src2)))], + IIC_SSE_UNPCK>; def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i128mem:$src2), !if(Is2Addr, @@ -3702,7 +4285,8 @@ multiclass sse2_unpack<bits<8> opc, string OpcodeStr, ValueType vt, !strconcat(OpcodeStr,"\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set VR128:$dst, (OpNode VR128:$src1, (bc_frag (memopv2i64 - addr:$src2))))]>; + addr:$src2))))], + IIC_SSE_UNPCK>; } multiclass sse2_unpack_y<bits<8> opc, string OpcodeStr, ValueType vt, @@ -3814,7 +4398,7 @@ multiclass sse2_pinsrw<bit Is2Addr = 1> { "pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}", "vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), [(set VR128:$dst, - (X86pinsrw VR128:$src1, GR32:$src2, imm:$src3))]>; + (X86pinsrw VR128:$src1, GR32:$src2, imm:$src3))], IIC_SSE_PINSRW>; def rmi : Ii8<0xC4, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i16mem:$src2, i32i8imm:$src3), @@ -3823,7 +4407,7 @@ multiclass sse2_pinsrw<bit Is2Addr = 1> { "vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), [(set VR128:$dst, (X86pinsrw VR128:$src1, (extloadi16 addr:$src2), - imm:$src3))]>; + imm:$src3))], IIC_SSE_PINSRW>; } // Extract @@ -3837,7 +4421,7 @@ def PEXTRWri : PDIi8<0xC5, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src1, i32i8imm:$src2), "pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}", [(set GR32:$dst, (X86pextrw (v8i16 VR128:$src1), - imm:$src2))]>; + imm:$src2))], IIC_SSE_PEXTRW>; // Insert let Predicates = [HasAVX] in { @@ -3861,9 +4445,10 @@ let ExeDomain = SSEPackedInt in { def VPMOVMSKBrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src), "pmovmskb\t{$src, $dst|$dst, $src}", - [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))]>, VEX; + [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))], + IIC_SSE_MOVMSK>, VEX; def VPMOVMSKBr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src), - "pmovmskb\t{$src, $dst|$dst, $src}", []>, VEX; + "pmovmskb\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK>, VEX; let Predicates = [HasAVX2] in { def VPMOVMSKBYrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR256:$src), @@ -3875,7 +4460,8 @@ def VPMOVMSKBYr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src), def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src), "pmovmskb\t{$src, $dst|$dst, $src}", - [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))]>; + [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))], + IIC_SSE_MOVMSK>; } // ExeDomain = SSEPackedInt @@ -3889,21 +4475,25 @@ let Uses = [EDI] in def VMASKMOVDQU : VPDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), "maskmovdqu\t{$mask, $src|$src, $mask}", - [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>, VEX; + [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)], + IIC_SSE_MASKMOV>, VEX; let Uses = [RDI] in def VMASKMOVDQU64 : VPDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), "maskmovdqu\t{$mask, $src|$src, $mask}", - [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)]>, VEX; + [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)], + IIC_SSE_MASKMOV>, VEX; let Uses = [EDI] in def MASKMOVDQU : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), "maskmovdqu\t{$mask, $src|$src, $mask}", - [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>; + [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)], + IIC_SSE_MASKMOV>; let Uses = [RDI] in def MASKMOVDQU64 : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), "maskmovdqu\t{$mask, $src|$src, $mask}", - [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)]>; + [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)], + IIC_SSE_MASKMOV>; } // ExeDomain = SSEPackedInt @@ -3917,54 +4507,65 @@ def MASKMOVDQU64 : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), def VMOVDI2PDIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v4i32 (scalar_to_vector GR32:$src)))]>, VEX; + (v4i32 (scalar_to_vector GR32:$src)))], IIC_SSE_MOVDQ>, + VEX; def VMOVDI2PDIrm : VPDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v4i32 (scalar_to_vector (loadi32 addr:$src))))]>, + (v4i32 (scalar_to_vector (loadi32 addr:$src))))], + IIC_SSE_MOVDQ>, VEX; def VMOV64toPQIrr : VRPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v2i64 (scalar_to_vector GR64:$src)))]>, VEX; + (v2i64 (scalar_to_vector GR64:$src)))], + IIC_SSE_MOVDQ>, VEX; def VMOV64toSDrr : VRPDI<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (bitconvert GR64:$src))]>, VEX; + [(set FR64:$dst, (bitconvert GR64:$src))], + IIC_SSE_MOVDQ>, VEX; def MOVDI2PDIrr : PDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v4i32 (scalar_to_vector GR32:$src)))]>; + (v4i32 (scalar_to_vector GR32:$src)))], IIC_SSE_MOVDQ>; def MOVDI2PDIrm : PDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v4i32 (scalar_to_vector (loadi32 addr:$src))))]>; + (v4i32 (scalar_to_vector (loadi32 addr:$src))))], + IIC_SSE_MOVDQ>; def MOV64toPQIrr : RPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v2i64 (scalar_to_vector GR64:$src)))]>; + (v2i64 (scalar_to_vector GR64:$src)))], + IIC_SSE_MOVDQ>; def MOV64toSDrr : RPDI<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (bitconvert GR64:$src))]>; + [(set FR64:$dst, (bitconvert GR64:$src))], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Move Int Doubleword to Single Scalar // def VMOVDI2SSrr : VPDI<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (bitconvert GR32:$src))]>, VEX; + [(set FR32:$dst, (bitconvert GR32:$src))], + IIC_SSE_MOVDQ>, VEX; def VMOVDI2SSrm : VPDI<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))]>, + [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))], + IIC_SSE_MOVDQ>, VEX; def MOVDI2SSrr : PDI<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (bitconvert GR32:$src))]>; + [(set FR32:$dst, (bitconvert GR32:$src))], + IIC_SSE_MOVDQ>; def MOVDI2SSrm : PDI<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", - [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))]>; + [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Move Packed Doubleword Int to Packed Double Int @@ -3972,20 +4573,22 @@ def MOVDI2SSrm : PDI<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src), def VMOVPDI2DIrr : VPDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR128:$src), "movd\t{$src, $dst|$dst, $src}", [(set GR32:$dst, (vector_extract (v4i32 VR128:$src), - (iPTR 0)))]>, VEX; + (iPTR 0)))], IIC_SSE_MOVD_ToGP>, VEX; def VMOVPDI2DImr : VPDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR128:$src), "movd\t{$src, $dst|$dst, $src}", [(store (i32 (vector_extract (v4i32 VR128:$src), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], IIC_SSE_MOVDQ>, + VEX; def MOVPDI2DIrr : PDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR128:$src), "movd\t{$src, $dst|$dst, $src}", [(set GR32:$dst, (vector_extract (v4i32 VR128:$src), - (iPTR 0)))]>; + (iPTR 0)))], IIC_SSE_MOVD_ToGP>; def MOVPDI2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR128:$src), "movd\t{$src, $dst|$dst, $src}", [(store (i32 (vector_extract (v4i32 VR128:$src), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Move Packed Doubleword Int first element to Doubleword Int @@ -3993,13 +4596,15 @@ def MOVPDI2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR128:$src), def VMOVPQIto64rr : I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", [(set GR64:$dst, (vector_extract (v2i64 VR128:$src), - (iPTR 0)))]>, + (iPTR 0)))], + IIC_SSE_MOVD_ToGP>, TB, OpSize, VEX, VEX_W, Requires<[HasAVX, In64BitMode]>; def MOVPQIto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", [(set GR64:$dst, (vector_extract (v2i64 VR128:$src), - (iPTR 0)))]>; + (iPTR 0)))], + IIC_SSE_MOVD_ToGP>; //===---------------------------------------------------------------------===// // Bitcast FR64 <-> GR64 @@ -4011,37 +4616,45 @@ def VMOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src), VEX; def VMOVSDto64rr : VRPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", - [(set GR64:$dst, (bitconvert FR64:$src))]>, VEX; + [(set GR64:$dst, (bitconvert FR64:$src))], + IIC_SSE_MOVDQ>, VEX; def VMOVSDto64mr : VRPDI<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src), "movq\t{$src, $dst|$dst, $src}", - [(store (i64 (bitconvert FR64:$src)), addr:$dst)]>, - VEX; + [(store (i64 (bitconvert FR64:$src)), addr:$dst)], + IIC_SSE_MOVDQ>, VEX; def MOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src), "movq\t{$src, $dst|$dst, $src}", - [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>; + [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))], + IIC_SSE_MOVDQ>; def MOVSDto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", - [(set GR64:$dst, (bitconvert FR64:$src))]>; + [(set GR64:$dst, (bitconvert FR64:$src))], + IIC_SSE_MOVD_ToGP>; def MOVSDto64mr : RPDI<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src), "movq\t{$src, $dst|$dst, $src}", - [(store (i64 (bitconvert FR64:$src)), addr:$dst)]>; + [(store (i64 (bitconvert FR64:$src)), addr:$dst)], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Move Scalar Single to Double Int // def VMOVSS2DIrr : VPDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(set GR32:$dst, (bitconvert FR32:$src))]>, VEX; + [(set GR32:$dst, (bitconvert FR32:$src))], + IIC_SSE_MOVD_ToGP>, VEX; def VMOVSS2DImr : VPDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(store (i32 (bitconvert FR32:$src)), addr:$dst)]>, VEX; + [(store (i32 (bitconvert FR32:$src)), addr:$dst)], + IIC_SSE_MOVDQ>, VEX; def MOVSS2DIrr : PDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(set GR32:$dst, (bitconvert FR32:$src))]>; + [(set GR32:$dst, (bitconvert FR32:$src))], + IIC_SSE_MOVD_ToGP>; def MOVSS2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src), "movd\t{$src, $dst|$dst, $src}", - [(store (i32 (bitconvert FR32:$src)), addr:$dst)]>; + [(store (i32 (bitconvert FR32:$src)), addr:$dst)], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Patterns and instructions to describe movd/movq to XMM register zero-extends @@ -4050,23 +4663,26 @@ let AddedComplexity = 15 in { def VMOVZDI2PDIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v4i32 (X86vzmovl - (v4i32 (scalar_to_vector GR32:$src)))))]>, - VEX; + (v4i32 (scalar_to_vector GR32:$src)))))], + IIC_SSE_MOVDQ>, VEX; def VMOVZQI2PQIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only [(set VR128:$dst, (v2i64 (X86vzmovl - (v2i64 (scalar_to_vector GR64:$src)))))]>, + (v2i64 (scalar_to_vector GR64:$src)))))], + IIC_SSE_MOVDQ>, VEX, VEX_W; } let AddedComplexity = 15 in { def MOVZDI2PDIrr : PDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v4i32 (X86vzmovl - (v4i32 (scalar_to_vector GR32:$src)))))]>; + (v4i32 (scalar_to_vector GR32:$src)))))], + IIC_SSE_MOVDQ>; def MOVZQI2PQIrr : RPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only [(set VR128:$dst, (v2i64 (X86vzmovl - (v2i64 (scalar_to_vector GR64:$src)))))]>; + (v2i64 (scalar_to_vector GR64:$src)))))], + IIC_SSE_MOVDQ>; } let AddedComplexity = 20 in { @@ -4074,13 +4690,14 @@ def VMOVZDI2PDIrm : VPDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v4i32 (X86vzmovl (v4i32 (scalar_to_vector - (loadi32 addr:$src))))))]>, - VEX; + (loadi32 addr:$src))))))], + IIC_SSE_MOVDQ>, VEX; def MOVZDI2PDIrm : PDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src), "movd\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v4i32 (X86vzmovl (v4i32 (scalar_to_vector - (loadi32 addr:$src))))))]>; + (loadi32 addr:$src))))))], + IIC_SSE_MOVDQ>; } let Predicates = [HasAVX] in { @@ -4138,7 +4755,8 @@ def VMOVQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), def MOVQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), "movq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, - (v2i64 (scalar_to_vector (loadi64 addr:$src))))]>, XS, + (v2i64 (scalar_to_vector (loadi64 addr:$src))))], + IIC_SSE_MOVDQ>, XS, Requires<[HasSSE2]>; // SSE2 instruction with XS Prefix //===---------------------------------------------------------------------===// @@ -4147,11 +4765,13 @@ def MOVQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), def VMOVPQI2QImr : VPDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src), "movq\t{$src, $dst|$dst, $src}", [(store (i64 (vector_extract (v2i64 VR128:$src), - (iPTR 0))), addr:$dst)]>, VEX; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOVDQ>, VEX; def MOVPQI2QImr : PDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src), "movq\t{$src, $dst|$dst, $src}", [(store (i64 (vector_extract (v2i64 VR128:$src), - (iPTR 0))), addr:$dst)]>; + (iPTR 0))), addr:$dst)], + IIC_SSE_MOVDQ>; //===---------------------------------------------------------------------===// // Store / copy lower 64-bits of a XMM register. @@ -4161,14 +4781,16 @@ def VMOVLQ128mr : VPDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src), [(int_x86_sse2_storel_dq addr:$dst, VR128:$src)]>, VEX; def MOVLQ128mr : PDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src), "movq\t{$src, $dst|$dst, $src}", - [(int_x86_sse2_storel_dq addr:$dst, VR128:$src)]>; + [(int_x86_sse2_storel_dq addr:$dst, VR128:$src)], + IIC_SSE_MOVDQ>; let AddedComplexity = 20 in def VMOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), "vmovq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 (scalar_to_vector - (loadi64 addr:$src))))))]>, + (loadi64 addr:$src))))))], + IIC_SSE_MOVDQ>, XS, VEX, Requires<[HasAVX]>; let AddedComplexity = 20 in @@ -4176,7 +4798,8 @@ def MOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), "movq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 (scalar_to_vector - (loadi64 addr:$src))))))]>, + (loadi64 addr:$src))))))], + IIC_SSE_MOVDQ>, XS, Requires<[HasSSE2]>; let Predicates = [HasAVX], AddedComplexity = 20 in { @@ -4208,25 +4831,29 @@ def : Pat<(v4i64 (X86vzload addr:$src)), let AddedComplexity = 15 in def VMOVZPQILo2PQIrr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vmovq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))]>, + [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))], + IIC_SSE_MOVQ_RR>, XS, VEX, Requires<[HasAVX]>; let AddedComplexity = 15 in def MOVZPQILo2PQIrr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "movq\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))]>, + [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))], + IIC_SSE_MOVQ_RR>, XS, Requires<[HasSSE2]>; let AddedComplexity = 20 in def VMOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "vmovq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v2i64 (X86vzmovl - (loadv2i64 addr:$src))))]>, + (loadv2i64 addr:$src))))], + IIC_SSE_MOVDQ>, XS, VEX, Requires<[HasAVX]>; let AddedComplexity = 20 in { def MOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "movq\t{$src, $dst|$dst, $src}", [(set VR128:$dst, (v2i64 (X86vzmovl - (loadv2i64 addr:$src))))]>, + (loadv2i64 addr:$src))))], + IIC_SSE_MOVDQ>, XS, Requires<[HasSSE2]>; } @@ -4247,12 +4874,15 @@ let AddedComplexity = 20 in { // Instructions to match in the assembler def VMOVQs64rr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src), - "movq\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W; + "movq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVDQ>, VEX, VEX_W; def VMOVQd64rr : VPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), - "movq\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W; + "movq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVDQ>, VEX, VEX_W; // Recognize "movd" with GR64 destination, but encode as a "movq" def VMOVQd64rr_alt : VPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), - "movd\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W; + "movd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_MOVDQ>, VEX, VEX_W; // Instructions for the disassembler // xr = XMM register @@ -4262,7 +4892,7 @@ let Predicates = [HasAVX] in def VMOVQxrxr: I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), "vmovq\t{$src, $dst|$dst, $src}", []>, VEX, XS; def MOVQxrxr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "movq\t{$src, $dst|$dst, $src}", []>, XS; + "movq\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVQ_RR>, XS; //===---------------------------------------------------------------------===// // SSE3 - Conversion Instructions @@ -4292,9 +4922,11 @@ def VCVTPD2DQYrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src), } def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtpd2dq\t{$src, $dst|$dst, $src}", []>; + "cvtpd2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>; def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtpd2dq\t{$src, $dst|$dst, $src}", []>; + "cvtpd2dq\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>; def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))), (VCVTTPD2DQYrr VR256:$src)>; @@ -4314,9 +4946,11 @@ def VCVTDQ2PDYrr : S3SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src), } def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src), - "cvtdq2pd\t{$src, $dst|$dst, $src}", []>; + "cvtdq2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RR>; def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), - "cvtdq2pd\t{$src, $dst|$dst, $src}", []>; + "cvtdq2pd\t{$src, $dst|$dst, $src}", [], + IIC_SSE_CVT_PD_RM>; // AVX 256-bit register conversion intrinsics def : Pat<(int_x86_avx_cvtdq2_pd_256 VR128:$src), @@ -4342,10 +4976,12 @@ multiclass sse3_replicate_sfp<bits<8> op, SDNode OpNode, string OpcodeStr, X86MemOperand x86memop> { def rr : S3SI<op, MRMSrcReg, (outs RC:$dst), (ins RC:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), - [(set RC:$dst, (vt (OpNode RC:$src)))]>; + [(set RC:$dst, (vt (OpNode RC:$src)))], + IIC_SSE_MOV_LH>; def rm : S3SI<op, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), - [(set RC:$dst, (OpNode (mem_frag addr:$src)))]>; + [(set RC:$dst, (OpNode (mem_frag addr:$src)))], + IIC_SSE_MOV_LH>; } let Predicates = [HasAVX] in { @@ -4401,12 +5037,13 @@ multiclass sse3_replicate_dfp<string OpcodeStr> { let neverHasSideEffects = 1 in def rr : S3DI<0x12, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), - []>; + [], IIC_SSE_MOV_LH>; def rm : S3DI<0x12, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), [(set VR128:$dst, (v2f64 (X86Movddup - (scalar_to_vector (loadf64 addr:$src)))))]>; + (scalar_to_vector (loadf64 addr:$src)))))], + IIC_SSE_MOV_LH>; } // FIXME: Merge with above classe when there're patterns for the ymm version @@ -4476,49 +5113,51 @@ let Predicates = [HasAVX] in { } def LDDQUrm : S3DI<0xF0, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src), "lddqu\t{$src, $dst|$dst, $src}", - [(set VR128:$dst, (int_x86_sse3_ldu_dq addr:$src))]>; + [(set VR128:$dst, (int_x86_sse3_ldu_dq addr:$src))], + IIC_SSE_LDDQU>; //===---------------------------------------------------------------------===// // SSE3 - Arithmetic //===---------------------------------------------------------------------===// multiclass sse3_addsub<Intrinsic Int, string OpcodeStr, RegisterClass RC, - X86MemOperand x86memop, bit Is2Addr = 1> { + X86MemOperand x86memop, OpndItins itins, + bit Is2Addr = 1> { def rr : I<0xD0, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (Int RC:$src1, RC:$src2))]>; + [(set RC:$dst, (Int RC:$src1, RC:$src2))], itins.rr>; def rm : I<0xD0, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (Int RC:$src1, (memop addr:$src2)))]>; + [(set RC:$dst, (Int RC:$src1, (memop addr:$src2)))], itins.rr>; } let Predicates = [HasAVX] in { let ExeDomain = SSEPackedSingle in { defm VADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "vaddsubps", VR128, - f128mem, 0>, TB, XD, VEX_4V; + f128mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V; defm VADDSUBPSY : sse3_addsub<int_x86_avx_addsub_ps_256, "vaddsubps", VR256, - f256mem, 0>, TB, XD, VEX_4V; + f256mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V; } let ExeDomain = SSEPackedDouble in { defm VADDSUBPD : sse3_addsub<int_x86_sse3_addsub_pd, "vaddsubpd", VR128, - f128mem, 0>, TB, OpSize, VEX_4V; + f128mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V; defm VADDSUBPDY : sse3_addsub<int_x86_avx_addsub_pd_256, "vaddsubpd", VR256, - f256mem, 0>, TB, OpSize, VEX_4V; + f256mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V; } } let Constraints = "$src1 = $dst", Predicates = [HasSSE3] in { let ExeDomain = SSEPackedSingle in defm ADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "addsubps", VR128, - f128mem>, TB, XD; + f128mem, SSE_ALU_F32P>, TB, XD; let ExeDomain = SSEPackedDouble in defm ADDSUBPD : sse3_addsub<int_x86_sse3_addsub_pd, "addsubpd", VR128, - f128mem>, TB, OpSize; + f128mem, SSE_ALU_F64P>, TB, OpSize; } //===---------------------------------------------------------------------===// @@ -4532,13 +5171,14 @@ multiclass S3D_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC, !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>; + [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_SSE_HADDSUB_RR>; def rm : S3DI<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))]>; + [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))], + IIC_SSE_HADDSUB_RM>; } multiclass S3_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC, X86MemOperand x86memop, SDNode OpNode, bit Is2Addr = 1> { @@ -4546,13 +5186,14 @@ multiclass S3_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC, !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>; + [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_SSE_HADDSUB_RR>; def rm : S3I<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))]>; + [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))], + IIC_SSE_HADDSUB_RM>; } let Predicates = [HasAVX] in { @@ -4600,7 +5241,7 @@ multiclass SS3I_unop_rm_int<bits<8> opc, string OpcodeStr, def rr128 : SS38I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src), !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), - [(set VR128:$dst, (IntId128 VR128:$src))]>, + [(set VR128:$dst, (IntId128 VR128:$src))], IIC_SSE_PABS_RR>, OpSize; def rm128 : SS38I<opc, MRMSrcMem, (outs VR128:$dst), @@ -4608,7 +5249,8 @@ multiclass SS3I_unop_rm_int<bits<8> opc, string OpcodeStr, !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), [(set VR128:$dst, (IntId128 - (bitconvert (memopv2i64 addr:$src))))]>, OpSize; + (bitconvert (memopv2i64 addr:$src))))], IIC_SSE_PABS_RM>, + OpSize; } /// SS3I_unop_rm_int_y - Simple SSSE3 unary op whose type can be v*{i8,i16,i32}. @@ -4657,17 +5299,37 @@ defm PABSD : SS3I_unop_rm_int<0x1E, "pabsd", // SSSE3 - Packed Binary Operator Instructions //===---------------------------------------------------------------------===// +def SSE_PHADDSUBD : OpndItins< + IIC_SSE_PHADDSUBD_RR, IIC_SSE_PHADDSUBD_RM +>; +def SSE_PHADDSUBSW : OpndItins< + IIC_SSE_PHADDSUBSW_RR, IIC_SSE_PHADDSUBSW_RM +>; +def SSE_PHADDSUBW : OpndItins< + IIC_SSE_PHADDSUBW_RR, IIC_SSE_PHADDSUBW_RM +>; +def SSE_PSHUFB : OpndItins< + IIC_SSE_PSHUFB_RR, IIC_SSE_PSHUFB_RM +>; +def SSE_PSIGN : OpndItins< + IIC_SSE_PSIGN_RR, IIC_SSE_PSIGN_RM +>; +def SSE_PMULHRSW : OpndItins< + IIC_SSE_PMULHRSW, IIC_SSE_PMULHRSW +>; + /// SS3I_binop_rm - Simple SSSE3 bin op multiclass SS3I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode, ValueType OpVT, RegisterClass RC, PatFrag memop_frag, - X86MemOperand x86memop, bit Is2Addr = 1> { + X86MemOperand x86memop, OpndItins itins, + bit Is2Addr = 1> { let isCommutable = 1 in def rr : SS38I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !if(Is2Addr, !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), - [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))]>, + [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))], itins.rr>, OpSize; def rm : SS38I<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2), @@ -4676,12 +5338,13 @@ multiclass SS3I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode, !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")), [(set RC:$dst, (OpVT (OpNode RC:$src1, - (bitconvert (memop_frag addr:$src2)))))]>, OpSize; + (bitconvert (memop_frag addr:$src2)))))], itins.rm>, OpSize; } /// SS3I_binop_rm_int - Simple SSSE3 bin op whose type can be v*{i8,i16,i32}. multiclass SS3I_binop_rm_int<bits<8> opc, string OpcodeStr, - Intrinsic IntId128, bit Is2Addr = 1> { + Intrinsic IntId128, OpndItins itins, + bit Is2Addr = 1> { let isCommutable = 1 in def rr128 : SS38I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2), @@ -4719,50 +5382,70 @@ multiclass SS3I_binop_rm_int_y<bits<8> opc, string OpcodeStr, let ImmT = NoImm, Predicates = [HasAVX] in { let isCommutable = 0 in { defm VPHADDW : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v8i16, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHADDD : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v4i32, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PHADDSUBD, 0>, VEX_4V; defm VPHSUBW : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v8i16, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHSUBD : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v4i32, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PHADDSUBD, 0>, VEX_4V; defm VPSIGNB : SS3I_binop_rm<0x08, "vpsignb", X86psign, v16i8, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PSIGN, 0>, VEX_4V; defm VPSIGNW : SS3I_binop_rm<0x09, "vpsignw", X86psign, v8i16, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PSIGN, 0>, VEX_4V; defm VPSIGND : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v4i32, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PSIGN, 0>, VEX_4V; defm VPSHUFB : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v16i8, VR128, - memopv2i64, i128mem, 0>, VEX_4V; + memopv2i64, i128mem, + SSE_PSHUFB, 0>, VEX_4V; defm VPHADDSW : SS3I_binop_rm_int<0x03, "vphaddsw", - int_x86_ssse3_phadd_sw_128, 0>, VEX_4V; + int_x86_ssse3_phadd_sw_128, + SSE_PHADDSUBSW, 0>, VEX_4V; defm VPHSUBSW : SS3I_binop_rm_int<0x07, "vphsubsw", - int_x86_ssse3_phsub_sw_128, 0>, VEX_4V; + int_x86_ssse3_phsub_sw_128, + SSE_PHADDSUBSW, 0>, VEX_4V; defm VPMADDUBSW : SS3I_binop_rm_int<0x04, "vpmaddubsw", - int_x86_ssse3_pmadd_ub_sw_128, 0>, VEX_4V; + int_x86_ssse3_pmadd_ub_sw_128, + SSE_PMADD, 0>, VEX_4V; } defm VPMULHRSW : SS3I_binop_rm_int<0x0B, "vpmulhrsw", - int_x86_ssse3_pmul_hr_sw_128, 0>, VEX_4V; + int_x86_ssse3_pmul_hr_sw_128, + SSE_PMULHRSW, 0>, VEX_4V; } let ImmT = NoImm, Predicates = [HasAVX2] in { let isCommutable = 0 in { defm VPHADDWY : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v16i16, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHADDDY : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v8i32, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHSUBWY : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v16i16, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHSUBDY : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v8i32, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPSIGNBY : SS3I_binop_rm<0x08, "vpsignb", X86psign, v32i8, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPSIGNWY : SS3I_binop_rm<0x09, "vpsignw", X86psign, v16i16, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPSIGNDY : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v8i32, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPSHUFBY : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v32i8, VR256, - memopv4i64, i256mem, 0>, VEX_4V; + memopv4i64, i256mem, + SSE_PHADDSUBW, 0>, VEX_4V; defm VPHADDSW : SS3I_binop_rm_int_y<0x03, "vphaddsw", int_x86_avx2_phadd_sw>, VEX_4V; defm VPHSUBSW : SS3I_binop_rm_int_y<0x07, "vphsubsw", @@ -4778,30 +5461,33 @@ defm VPMULHRSW : SS3I_binop_rm_int_y<0x0B, "vpmulhrsw", let ImmT = NoImm, Constraints = "$src1 = $dst" in { let isCommutable = 0 in { defm PHADDW : SS3I_binop_rm<0x01, "phaddw", X86hadd, v8i16, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PHADDSUBW>; defm PHADDD : SS3I_binop_rm<0x02, "phaddd", X86hadd, v4i32, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PHADDSUBD>; defm PHSUBW : SS3I_binop_rm<0x05, "phsubw", X86hsub, v8i16, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PHADDSUBW>; defm PHSUBD : SS3I_binop_rm<0x06, "phsubd", X86hsub, v4i32, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PHADDSUBD>; defm PSIGNB : SS3I_binop_rm<0x08, "psignb", X86psign, v16i8, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PSIGN>; defm PSIGNW : SS3I_binop_rm<0x09, "psignw", X86psign, v8i16, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PSIGN>; defm PSIGND : SS3I_binop_rm<0x0A, "psignd", X86psign, v4i32, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PSIGN>; defm PSHUFB : SS3I_binop_rm<0x00, "pshufb", X86pshufb, v16i8, VR128, - memopv2i64, i128mem>; + memopv2i64, i128mem, SSE_PSHUFB>; defm PHADDSW : SS3I_binop_rm_int<0x03, "phaddsw", - int_x86_ssse3_phadd_sw_128>; + int_x86_ssse3_phadd_sw_128, + SSE_PHADDSUBSW>; defm PHSUBSW : SS3I_binop_rm_int<0x07, "phsubsw", - int_x86_ssse3_phsub_sw_128>; + int_x86_ssse3_phsub_sw_128, + SSE_PHADDSUBSW>; defm PMADDUBSW : SS3I_binop_rm_int<0x04, "pmaddubsw", - int_x86_ssse3_pmadd_ub_sw_128>; + int_x86_ssse3_pmadd_ub_sw_128, SSE_PMADD>; } defm PMULHRSW : SS3I_binop_rm_int<0x0B, "pmulhrsw", - int_x86_ssse3_pmul_hr_sw_128>; + int_x86_ssse3_pmul_hr_sw_128, + SSE_PMULHRSW>; } //===---------------------------------------------------------------------===// @@ -4816,7 +5502,7 @@ multiclass ssse3_palign<string asm, bit Is2Addr = 1> { !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"), !strconcat(asm, "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")), - []>, OpSize; + [], IIC_SSE_PALIGNR>, OpSize; let mayLoad = 1 in def R128rm : SS3AI<0x0F, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i128mem:$src2, i8imm:$src3), @@ -4824,7 +5510,7 @@ multiclass ssse3_palign<string asm, bit Is2Addr = 1> { !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"), !strconcat(asm, "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")), - []>, OpSize; + [], IIC_SSE_PALIGNR>, OpSize; } } @@ -4898,11 +5584,11 @@ def MWAIT : PseudoI<(outs), (ins GR32:$src1, GR32:$src2), } let Uses = [EAX, ECX, EDX] in -def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", []>, TB, - Requires<[HasSSE3]>; +def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", [], IIC_SSE_MONITOR>, + TB, Requires<[HasSSE3]>; let Uses = [ECX, EAX] in -def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait", []>, TB, - Requires<[HasSSE3]>; +def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait", [], IIC_SSE_MWAIT>, + TB, Requires<[HasSSE3]>; def : InstAlias<"mwait %eax, %ecx", (MWAITrr)>, Requires<[In32BitMode]>; def : InstAlias<"mwait %rax, %rcx", (MWAITrr)>, Requires<[In64BitMode]>; diff --git a/lib/Target/X86/X86Schedule.td b/lib/Target/X86/X86Schedule.td index 39523e6..9692833 100644 --- a/lib/Target/X86/X86Schedule.td +++ b/lib/Target/X86/X86Schedule.td @@ -103,6 +103,142 @@ def IIC_CALL_FAR_PTR : InstrItinClass; // ret def IIC_RET : InstrItinClass; def IIC_RET_IMM : InstrItinClass; +// SSE scalar/parallel binary operations +def IIC_SSE_ALU_F32S_RR : InstrItinClass; +def IIC_SSE_ALU_F32S_RM : InstrItinClass; +def IIC_SSE_ALU_F64S_RR : InstrItinClass; +def IIC_SSE_ALU_F64S_RM : InstrItinClass; +def IIC_SSE_MUL_F32S_RR : InstrItinClass; +def IIC_SSE_MUL_F32S_RM : InstrItinClass; +def IIC_SSE_MUL_F64S_RR : InstrItinClass; +def IIC_SSE_MUL_F64S_RM : InstrItinClass; +def IIC_SSE_DIV_F32S_RR : InstrItinClass; +def IIC_SSE_DIV_F32S_RM : InstrItinClass; +def IIC_SSE_DIV_F64S_RR : InstrItinClass; +def IIC_SSE_DIV_F64S_RM : InstrItinClass; +def IIC_SSE_ALU_F32P_RR : InstrItinClass; +def IIC_SSE_ALU_F32P_RM : InstrItinClass; +def IIC_SSE_ALU_F64P_RR : InstrItinClass; +def IIC_SSE_ALU_F64P_RM : InstrItinClass; +def IIC_SSE_MUL_F32P_RR : InstrItinClass; +def IIC_SSE_MUL_F32P_RM : InstrItinClass; +def IIC_SSE_MUL_F64P_RR : InstrItinClass; +def IIC_SSE_MUL_F64P_RM : InstrItinClass; +def IIC_SSE_DIV_F32P_RR : InstrItinClass; +def IIC_SSE_DIV_F32P_RM : InstrItinClass; +def IIC_SSE_DIV_F64P_RR : InstrItinClass; +def IIC_SSE_DIV_F64P_RM : InstrItinClass; + +def IIC_SSE_COMIS_RR : InstrItinClass; +def IIC_SSE_COMIS_RM : InstrItinClass; + +def IIC_SSE_HADDSUB_RR : InstrItinClass; +def IIC_SSE_HADDSUB_RM : InstrItinClass; + +def IIC_SSE_BIT_P_RR : InstrItinClass; +def IIC_SSE_BIT_P_RM : InstrItinClass; + +def IIC_SSE_INTALU_P_RR : InstrItinClass; +def IIC_SSE_INTALU_P_RM : InstrItinClass; +def IIC_SSE_INTALUQ_P_RR : InstrItinClass; +def IIC_SSE_INTALUQ_P_RM : InstrItinClass; + +def IIC_SSE_INTMUL_P_RR : InstrItinClass; +def IIC_SSE_INTMUL_P_RM : InstrItinClass; + +def IIC_SSE_INTSH_P_RR : InstrItinClass; +def IIC_SSE_INTSH_P_RM : InstrItinClass; +def IIC_SSE_INTSH_P_RI : InstrItinClass; + +def IIC_SSE_CMPP_RR : InstrItinClass; +def IIC_SSE_CMPP_RM : InstrItinClass; + +def IIC_SSE_SHUFP : InstrItinClass; +def IIC_SSE_PSHUF : InstrItinClass; + +def IIC_SSE_UNPCK : InstrItinClass; + +def IIC_SSE_MOVMSK : InstrItinClass; +def IIC_SSE_MASKMOV : InstrItinClass; + +def IIC_SSE_PEXTRW : InstrItinClass; +def IIC_SSE_PINSRW : InstrItinClass; + +def IIC_SSE_PABS_RR : InstrItinClass; +def IIC_SSE_PABS_RM : InstrItinClass; + +def IIC_SSE_SQRTP_RR : InstrItinClass; +def IIC_SSE_SQRTP_RM : InstrItinClass; +def IIC_SSE_SQRTS_RR : InstrItinClass; +def IIC_SSE_SQRTS_RM : InstrItinClass; + +def IIC_SSE_RCPP_RR : InstrItinClass; +def IIC_SSE_RCPP_RM : InstrItinClass; +def IIC_SSE_RCPS_RR : InstrItinClass; +def IIC_SSE_RCPS_RM : InstrItinClass; + +def IIC_SSE_MOV_S_RR : InstrItinClass; +def IIC_SSE_MOV_S_RM : InstrItinClass; +def IIC_SSE_MOV_S_MR : InstrItinClass; + +def IIC_SSE_MOVA_P_RR : InstrItinClass; +def IIC_SSE_MOVA_P_RM : InstrItinClass; +def IIC_SSE_MOVA_P_MR : InstrItinClass; + +def IIC_SSE_MOVU_P_RR : InstrItinClass; +def IIC_SSE_MOVU_P_RM : InstrItinClass; +def IIC_SSE_MOVU_P_MR : InstrItinClass; + +def IIC_SSE_MOVDQ : InstrItinClass; +def IIC_SSE_MOVD_ToGP : InstrItinClass; +def IIC_SSE_MOVQ_RR : InstrItinClass; + +def IIC_SSE_MOV_LH : InstrItinClass; + +def IIC_SSE_LDDQU : InstrItinClass; + +def IIC_SSE_MOVNT : InstrItinClass; + +def IIC_SSE_PHADDSUBD_RR : InstrItinClass; +def IIC_SSE_PHADDSUBD_RM : InstrItinClass; +def IIC_SSE_PHADDSUBSW_RR : InstrItinClass; +def IIC_SSE_PHADDSUBSW_RM : InstrItinClass; +def IIC_SSE_PHADDSUBW_RR : InstrItinClass; +def IIC_SSE_PHADDSUBW_RM : InstrItinClass; +def IIC_SSE_PSHUFB_RR : InstrItinClass; +def IIC_SSE_PSHUFB_RM : InstrItinClass; +def IIC_SSE_PSIGN_RR : InstrItinClass; +def IIC_SSE_PSIGN_RM : InstrItinClass; + +def IIC_SSE_PMADD : InstrItinClass; +def IIC_SSE_PMULHRSW : InstrItinClass; +def IIC_SSE_PALIGNR : InstrItinClass; +def IIC_SSE_MWAIT : InstrItinClass; +def IIC_SSE_MONITOR : InstrItinClass; + +def IIC_SSE_PREFETCH : InstrItinClass; +def IIC_SSE_PAUSE : InstrItinClass; +def IIC_SSE_LFENCE : InstrItinClass; +def IIC_SSE_MFENCE : InstrItinClass; +def IIC_SSE_SFENCE : InstrItinClass; +def IIC_SSE_LDMXCSR : InstrItinClass; +def IIC_SSE_STMXCSR : InstrItinClass; + +def IIC_SSE_CVT_PD_RR : InstrItinClass; +def IIC_SSE_CVT_PD_RM : InstrItinClass; +def IIC_SSE_CVT_PS_RR : InstrItinClass; +def IIC_SSE_CVT_PS_RM : InstrItinClass; +def IIC_SSE_CVT_PI2PS_RR : InstrItinClass; +def IIC_SSE_CVT_PI2PS_RM : InstrItinClass; +def IIC_SSE_CVT_Scalar_RR : InstrItinClass; +def IIC_SSE_CVT_Scalar_RM : InstrItinClass; +def IIC_SSE_CVT_SS2SI32_RM : InstrItinClass; +def IIC_SSE_CVT_SS2SI32_RR : InstrItinClass; +def IIC_SSE_CVT_SS2SI64_RM : InstrItinClass; +def IIC_SSE_CVT_SS2SI64_RR : InstrItinClass; +def IIC_SSE_CVT_SD2SI_RM : InstrItinClass; +def IIC_SSE_CVT_SD2SI_RR : InstrItinClass; + //===----------------------------------------------------------------------===// // Processor instruction itineraries. diff --git a/lib/Target/X86/X86ScheduleAtom.td b/lib/Target/X86/X86ScheduleAtom.td index 55a9199..67cd0f2 100644 --- a/lib/Target/X86/X86ScheduleAtom.td +++ b/lib/Target/X86/X86ScheduleAtom.td @@ -125,12 +125,159 @@ def AtomItineraries : ProcessorItineraries< InstrItinData<IIC_LOOPE, [InstrStage<8, [Port0, Port1]>] >, InstrItinData<IIC_LOOPNE, [InstrStage<17, [Port0, Port1]>] >, // call - all but reg/imm - InstrItinData<IIC_CALL_RI, [InstrStage<1, [Port0], 0>, InstrStage<1, [Port1]>] >, + InstrItinData<IIC_CALL_RI, [InstrStage<1, [Port0], 0>, + InstrStage<1, [Port1]>] >, InstrItinData<IIC_CALL_MEM, [InstrStage<15, [Port0, Port1]>] >, InstrItinData<IIC_CALL_FAR_MEM, [InstrStage<40, [Port0, Port1]>] >, InstrItinData<IIC_CALL_FAR_PTR, [InstrStage<39, [Port0, Port1]>] >, //ret InstrItinData<IIC_RET, [InstrStage<79, [Port0, Port1]>] >, - InstrItinData<IIC_RET_IMM, [InstrStage<1, [Port0], 0>, InstrStage<1, [Port1]>] > + InstrItinData<IIC_RET_IMM, [InstrStage<1, [Port0], 0>, InstrStage<1, [Port1]>] >, + // SSE binary operations + // arithmetic fp scalar + InstrItinData<IIC_SSE_ALU_F32S_RR, [InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_ALU_F32S_RM, [InstrStage<5, [Port0], 0>, + InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_ALU_F64S_RR, [InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_ALU_F64S_RM, [InstrStage<5, [Port0], 0>, + InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_MUL_F32S_RR, [InstrStage<4, [Port0]>] >, + InstrItinData<IIC_SSE_MUL_F32S_RM, [InstrStage<4, [Port0]>] >, + InstrItinData<IIC_SSE_MUL_F64S_RR, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_MUL_F64S_RM, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_DIV_F32S_RR, [InstrStage<34, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F32S_RM, [InstrStage<34, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F64S_RR, [InstrStage<62, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F64S_RM, [InstrStage<62, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_COMIS_RR, [InstrStage<9, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_COMIS_RM, [InstrStage<10, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_HADDSUB_RR, [InstrStage<8, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_HADDSUB_RM, [InstrStage<9, [Port0, Port1]>] >, + + // arithmetic fp parallel + InstrItinData<IIC_SSE_ALU_F32P_RR, [InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_ALU_F32P_RM, [InstrStage<5, [Port0], 0>, + InstrStage<5, [Port1]>] >, + InstrItinData<IIC_SSE_ALU_F64P_RR, [InstrStage<6, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_ALU_F64P_RM, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MUL_F32P_RR, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_MUL_F32P_RM, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_MUL_F64P_RR, [InstrStage<9, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MUL_F64P_RM, [InstrStage<10, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F32P_RR, [InstrStage<70, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F32P_RM, [InstrStage<70, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F64P_RR, [InstrStage<125, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_DIV_F64P_RM, [InstrStage<125, [Port0, Port1]>] >, + + // bitwise parallel + InstrItinData<IIC_SSE_BIT_P_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_BIT_P_RM, [InstrStage<1, [Port0]>] >, + + // arithmetic int parallel + InstrItinData<IIC_SSE_INTALU_P_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_INTALU_P_RM, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_INTALUQ_P_RR, [InstrStage<2, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_INTALUQ_P_RM, [InstrStage<3, [Port0, Port1]>] >, + + // multiply int parallel + InstrItinData<IIC_SSE_INTMUL_P_RR, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_INTMUL_P_RM, [InstrStage<5, [Port0]>] >, + + // shift parallel + InstrItinData<IIC_SSE_INTSH_P_RR, [InstrStage<2, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_INTSH_P_RM, [InstrStage<3, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_INTSH_P_RI, [InstrStage<1, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_CMPP_RR, [InstrStage<6, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CMPP_RM, [InstrStage<7, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_SHUFP, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_PSHUF, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_UNPCK, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_SQRTP_RR, [InstrStage<13, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_SQRTP_RM, [InstrStage<14, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_SQRTS_RR, [InstrStage<11, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_SQRTS_RM, [InstrStage<12, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_RCPP_RR, [InstrStage<9, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_RCPP_RM, [InstrStage<10, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_RCPS_RR, [InstrStage<4, [Port0]>] >, + InstrItinData<IIC_SSE_RCPS_RM, [InstrStage<4, [Port0]>] >, + + InstrItinData<IIC_SSE_MOVMSK, [InstrStage<3, [Port0]>] >, + InstrItinData<IIC_SSE_MASKMOV, [InstrStage<2, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_PEXTRW, [InstrStage<4, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PINSRW, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_PABS_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PABS_RM, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_MOV_S_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MOV_S_RM, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_MOV_S_MR, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_MOVA_P_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MOVA_P_RM, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_MOVA_P_MR, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_MOVU_P_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MOVU_P_RM, [InstrStage<3, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MOVU_P_MR, [InstrStage<2, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_MOV_LH, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_LDDQU, [InstrStage<3, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_MOVDQ, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_MOVD_ToGP, [InstrStage<3, [Port0]>] >, + InstrItinData<IIC_SSE_MOVQ_RR, [InstrStage<1, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_MOVNT, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_PREFETCH, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_PAUSE, [InstrStage<17, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_LFENCE, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MFENCE, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_SFENCE, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_LDMXCSR, [InstrStage<5, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_STMXCSR, [InstrStage<15, [Port0, Port1]>] >, + + InstrItinData<IIC_SSE_PHADDSUBD_RR, [InstrStage<3, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PHADDSUBD_RM, [InstrStage<4, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PHADDSUBSW_RR, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PHADDSUBSW_RM, [InstrStage<8, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PHADDSUBW_RR, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PHADDSUBW_RM, [InstrStage<8, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PSHUFB_RR, [InstrStage<4, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PSHUFB_RM, [InstrStage<5, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PSIGN_RR, [InstrStage<1, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_PSIGN_RM, [InstrStage<1, [Port0]>] >, + + InstrItinData<IIC_SSE_PMADD, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_PMULHRSW, [InstrStage<5, [Port0]>] >, + InstrItinData<IIC_SSE_PALIGNR, [InstrStage<1, [Port0]>] >, + InstrItinData<IIC_SSE_MWAIT, [InstrStage<46, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_MONITOR, [InstrStage<45, [Port0, Port1]>] >, + + // conversions + // to/from PD ... + InstrItinData<IIC_SSE_CVT_PD_RR, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_PD_RM, [InstrStage<8, [Port0, Port1]>] >, + // to/from PS except to/from PD and PS2PI + InstrItinData<IIC_SSE_CVT_PS_RR, [InstrStage<6, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_PS_RM, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_Scalar_RR, [InstrStage<6, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_Scalar_RM, [InstrStage<7, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SS2SI32_RR, [InstrStage<8, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SS2SI32_RM, [InstrStage<9, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SS2SI64_RR, [InstrStage<9, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SS2SI64_RM, [InstrStage<10, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SD2SI_RR, [InstrStage<8, [Port0, Port1]>] >, + InstrItinData<IIC_SSE_CVT_SD2SI_RM, [InstrStage<9, [Port0, Port1]>] > ]>; |