X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FX86%2FX86InstrMMX.td;h=d484695f60ea2214e70dd6e154cd57e2cdcb8f7b;hb=22b942aa4df824adbd3f6eaede53abe451f616e9;hp=b301b405b6bd5ae4a6ac1f910492614b241421db;hpb=c32a7f98ab318b5df0a3239cd3c24ab7dbf4e720;p=oota-llvm.git diff --git a/lib/Target/X86/X86InstrMMX.td b/lib/Target/X86/X86InstrMMX.td index b301b405b6b..d484695f60e 100644 --- a/lib/Target/X86/X86InstrMMX.td +++ b/lib/Target/X86/X86InstrMMX.td @@ -2,8 +2,8 @@ // // The LLVM Compiler Infrastructure // -// This file was developed by the Evan Cheng and is distributed under -// the University of Illinois Open Source License. See LICENSE.TXT for details. +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // @@ -14,103 +14,506 @@ //===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===// -// Instruction templates +// MMX Pattern Fragments +//===----------------------------------------------------------------------===// + +def load_mmx : PatFrag<(ops node:$ptr), (v1i64 (load node:$ptr))>; + +def bc_v8i8 : PatFrag<(ops node:$in), (v8i8 (bitconvert node:$in))>; +def bc_v4i16 : PatFrag<(ops node:$in), (v4i16 (bitconvert node:$in))>; +def bc_v2i32 : PatFrag<(ops node:$in), (v2i32 (bitconvert node:$in))>; +def bc_v1i64 : PatFrag<(ops node:$in), (v1i64 (bitconvert node:$in))>; + //===----------------------------------------------------------------------===// +// MMX Masks +//===----------------------------------------------------------------------===// + +// MMX_SHUFFLE_get_shuf_imm xform function: convert vector_shuffle mask to +// PSHUFW imm. +def MMX_SHUFFLE_get_shuf_imm : SDNodeXForm; + +// Patterns for: vector_shuffle v1, v2, <2, 6, 3, 7, ...> +def MMX_UNPCKH_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isUNPCKHMask(N); +}]>; + +// Patterns for: vector_shuffle v1, v2, <0, 4, 2, 5, ...> +def MMX_UNPCKL_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isUNPCKLMask(N); +}]>; + +// Patterns for: vector_shuffle v1, , <0, 0, 1, 1, ...> +def MMX_UNPCKH_v_undef_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isUNPCKH_v_undef_Mask(N); +}]>; -// MMXI - MMX instructions with TB prefix. -// MMX2I - MMX / SSE2 instructions with TB and OpSize prefixes. -// MMXIi8 - MMX instructions with ImmT == Imm8 and TB prefix. -class MMXI o, Format F, dag ops, string asm, list pattern> - : I, TB, Requires<[HasMMX]>; -class MMX2I o, Format F, dag ops, string asm, list pattern> - : I, TB, OpSize, Requires<[HasSSE2]>; -class MMXIi8 o, Format F, dag ops, string asm, list pattern> - : Ii8, TB, Requires<[HasMMX]>; +// Patterns for: vector_shuffle v1, , <2, 2, 3, 3, ...> +def MMX_UNPCKL_v_undef_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isUNPCKL_v_undef_Mask(N); +}]>; -// Some 'special' instructions -def IMPLICIT_DEF_VR64 : I<0, Pseudo, (ops VR64:$dst), - "#IMPLICIT_DEF $dst", - [(set VR64:$dst, (v8i8 (undef)))]>, - Requires<[HasMMX]>; +// Patterns for shuffling. +def MMX_PSHUFW_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isPSHUFDMask(N); +}], MMX_SHUFFLE_get_shuf_imm>; -// 64-bit vector undef's. -def : Pat<(v8i8 (undef)), (IMPLICIT_DEF_VR64)>; -def : Pat<(v4i16 (undef)), (IMPLICIT_DEF_VR64)>; -def : Pat<(v2i32 (undef)), (IMPLICIT_DEF_VR64)>; +// Patterns for: vector_shuffle v1, v2, <4, 5, 2, 3>; etc. +def MMX_MOVL_shuffle_mask : PatLeaf<(build_vector), [{ + return X86::isMOVLMask(N); +}]>; //===----------------------------------------------------------------------===// -// MMX Pattern Fragments +// MMX Multiclasses //===----------------------------------------------------------------------===// -def loadv2i32 : PatFrag<(ops node:$ptr), (v2i32 (load node:$ptr))>; +let isTwoAddress = 1 in { + // MMXI_binop_rm - Simple MMX binary operator. + multiclass MMXI_binop_rm opc, string OpcodeStr, SDNode OpNode, + ValueType OpVT, bit Commutable = 0> { + def rr : MMXI { + let isCommutable = Commutable; + } + def rm : MMXI; + } + + multiclass MMXI_binop_rm_int opc, string OpcodeStr, Intrinsic IntId, + bit Commutable = 0> { + def rr : MMXI { + let isCommutable = Commutable; + } + def rm : MMXI; + } + + // MMXI_binop_rm_v1i64 - Simple MMX binary operator whose type is v1i64. + // + // FIXME: we could eliminate this and use MMXI_binop_rm instead if tblgen knew + // to collapse (bitconvert VT to VT) into its operand. + // + multiclass MMXI_binop_rm_v1i64 opc, string OpcodeStr, SDNode OpNode, + bit Commutable = 0> { + def rr : MMXI { + let isCommutable = Commutable; + } + def rm : MMXI; + } + + multiclass MMXI_binop_rmi_int opc, bits<8> opc2, Format ImmForm, + string OpcodeStr, Intrinsic IntId, + Intrinsic IntId2> { + def rr : MMXI; + def rm : MMXI; + def ri : MMXIi8; + } +} //===----------------------------------------------------------------------===// -// MMX EMMS Instruction +// MMX EMMS & FEMMS Instructions //===----------------------------------------------------------------------===// -def EMMS : MMXI<0x77, RawFrm, (ops), "emms", [(int_x86_mmx_emms)]>; +def MMX_EMMS : MMXI<0x77, RawFrm, (outs), (ins), "emms", [(int_x86_mmx_emms)]>; +def MMX_FEMMS : MMXI<0x0E, RawFrm, (outs), (ins), "femms", [(int_x86_mmx_femms)]>; //===----------------------------------------------------------------------===// // MMX Scalar Instructions //===----------------------------------------------------------------------===// -// Move Instructions -def MOVD64rr : MMXI<0x6E, MRMSrcReg, (ops VR64:$dst, GR32:$src), - "movd {$src, $dst|$dst, $src}", []>; -def MOVD64rm : MMXI<0x6E, MRMSrcMem, (ops VR64:$dst, i32mem:$src), - "movd {$src, $dst|$dst, $src}", []>; -def MOVD64mr : MMXI<0x7E, MRMDestMem, (ops i32mem:$dst, VR64:$src), - "movd {$src, $dst|$dst, $src}", []>; - -def MOVQ64rr : MMXI<0x6F, MRMSrcReg, (ops VR64:$dst, VR64:$src), - "movq {$src, $dst|$dst, $src}", []>; -def MOVQ64rm : MMXI<0x6F, MRMSrcMem, (ops VR64:$dst, i64mem:$src), - "movq {$src, $dst|$dst, $src}", - [(set VR64:$dst, (loadv2i32 addr:$src))]>; -def MOVQ64mr : MMXI<0x7F, MRMDestMem, (ops i64mem:$dst, VR64:$src), - "movq {$src, $dst|$dst, $src}", - [(store (v2i32 VR64:$src), addr:$dst)]>; - -// Conversion instructions -def CVTPI2PSrr : MMXI<0x2A, MRMSrcReg, (ops VR128:$dst, VR64:$src), - "cvtpi2ps {$src, $dst|$dst, $src}", []>; -def CVTPI2PSrm : MMXI<0x2A, MRMSrcMem, (ops VR128:$dst, i64mem:$src), - "cvtpi2ps {$src, $dst|$dst, $src}", []>; -def CVTPI2PDrr : MMX2I<0x2A, MRMSrcReg, (ops VR128:$dst, VR64:$src), - "cvtpi2pd {$src, $dst|$dst, $src}", []>; -def CVTPI2PDrm : MMX2I<0x2A, MRMSrcMem, (ops VR128:$dst, i64mem:$src), - "cvtpi2pd {$src, $dst|$dst, $src}", []>; -def CVTTPS2PIrr: I<0x2C, MRMSrcReg, (ops VR64:$dst, VR128:$src), - "cvttps2pi {$src, $dst|$dst, $src}", []>, TB, - Requires<[HasSSE2]>; -def CVTTPS2PIrm: I<0x2C, MRMSrcMem, (ops VR64:$dst, f64mem:$src), - "cvttps2pi {$src, $dst|$dst, $src}", []>, TB, - Requires<[HasMMX]>; -def CVTPS2PIrr : MMXI<0x2D, MRMSrcReg, (ops VR64:$dst, VR128:$src), - "cvtps2pi {$src, $dst|$dst, $src}", []>; -def CVTPS2PIrm : MMXI<0x2D, MRMSrcMem, (ops VR64:$dst, f64mem:$src), - "cvtps2pi {$src, $dst|$dst, $src}", []>; -def CVTPD2PIrr : MMX2I<0x2D, MRMSrcReg, (ops VR64:$dst, VR128:$src), - "cvtpd2pi {$src, $dst|$dst, $src}", []>; -def CVTPD2PIrm : MMX2I<0x2D, MRMSrcMem, (ops VR64:$dst, f128mem:$src), - "cvtpd2pi {$src, $dst|$dst, $src}", []>; - -// Shuffle and unpack instructions -def PSHUFWri : MMXIi8<0x70, MRMSrcReg, - (ops VR64:$dst, VR64:$src1, i8imm:$src2), - "pshufw {$src2, $src1, $dst|$dst, $src1, $src2}", []>; -def PSHUFWmi : MMXIi8<0x70, MRMSrcMem, - (ops VR64:$dst, i64mem:$src1, i8imm:$src2), - "pshufw {$src2, $src1, $dst|$dst, $src1, $src2}", []>; +// Data Transfer Instructions +def MMX_MOVD64rr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src), + "movd\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, (v2i32 (scalar_to_vector GR32:$src)))]>; +let isSimpleLoad = 1, isReMaterializable = 1 in +def MMX_MOVD64rm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src), + "movd\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, (v2i32 (scalar_to_vector (loadi32 addr:$src))))]>; +let mayStore = 1 in +def MMX_MOVD64mr : MMXI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR64:$src), + "movd\t{$src, $dst|$dst, $src}", []>; + +let neverHasSideEffects = 1 in +def MMX_MOVD64to64rr : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src), + "movd\t{$src, $dst|$dst, $src}", []>; + +let neverHasSideEffects = 1 in +def MMX_MOVD64from64rr : MMXRI<0x7E, MRMSrcReg, (outs GR64:$dst), (ins VR64:$src), + "movd\t{$src, $dst|$dst, $src}", []>; + +let neverHasSideEffects = 1 in +def MMX_MOVQ64rr : MMXI<0x6F, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src), + "movq\t{$src, $dst|$dst, $src}", []>; +let isSimpleLoad = 1, isReMaterializable = 1, mayHaveSideEffects = 1 in +def MMX_MOVQ64rm : MMXI<0x6F, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src), + "movq\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, (load_mmx addr:$src))]>; +def MMX_MOVQ64mr : MMXI<0x7F, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src), + "movq\t{$src, $dst|$dst, $src}", + [(store (v1i64 VR64:$src), addr:$dst)]>; + +def MMX_MOVDQ2Qrr : MMXID<0xD6, MRMDestMem, (outs VR64:$dst), (ins VR128:$src), + "movdq2q\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, + (v1i64 (bitconvert + (i64 (vector_extract (v2i64 VR128:$src), + (iPTR 0))))))]>; + +def MMX_MOVQ2DQrr : MMXIS<0xD6, MRMDestMem, (outs VR128:$dst), (ins VR64:$src), + "movq2dq\t{$src, $dst|$dst, $src}", + [(set VR128:$dst, + (v2i64 (vector_shuffle immAllZerosV, + (v2i64 (scalar_to_vector (i64 (bitconvert VR64:$src)))), + MOVL_shuffle_mask)))]>; + +def MMX_MOVNTQmr : MMXI<0xE7, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src), + "movntq\t{$src, $dst|$dst, $src}", + [(int_x86_mmx_movnt_dq addr:$dst, VR64:$src)]>; + +let AddedComplexity = 15 in +// movd to MMX register zero-extends +def MMX_MOVZDI2PDIrr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src), + "movd\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, + (v2i32 (vector_shuffle immAllZerosV, + (v2i32 (scalar_to_vector GR32:$src)), + MMX_MOVL_shuffle_mask)))]>; +let AddedComplexity = 20 in +def MMX_MOVZDI2PDIrm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src), + "movd\t{$src, $dst|$dst, $src}", + [(set VR64:$dst, + (v2i32 (vector_shuffle immAllZerosV, + (v2i32 (scalar_to_vector + (loadi32 addr:$src))), + MMX_MOVL_shuffle_mask)))]>; + +// Arithmetic Instructions + +// -- Addition +defm MMX_PADDB : MMXI_binop_rm<0xFC, "paddb", add, v8i8, 1>; +defm MMX_PADDW : MMXI_binop_rm<0xFD, "paddw", add, v4i16, 1>; +defm MMX_PADDD : MMXI_binop_rm<0xFE, "paddd", add, v2i32, 1>; +defm MMX_PADDQ : MMXI_binop_rm<0xD4, "paddq", add, v1i64, 1>; + +defm MMX_PADDSB : MMXI_binop_rm_int<0xEC, "paddsb" , int_x86_mmx_padds_b, 1>; +defm MMX_PADDSW : MMXI_binop_rm_int<0xED, "paddsw" , int_x86_mmx_padds_w, 1>; + +defm MMX_PADDUSB : MMXI_binop_rm_int<0xDC, "paddusb", int_x86_mmx_paddus_b, 1>; +defm MMX_PADDUSW : MMXI_binop_rm_int<0xDD, "paddusw", int_x86_mmx_paddus_w, 1>; + +// -- Subtraction +defm MMX_PSUBB : MMXI_binop_rm<0xF8, "psubb", sub, v8i8>; +defm MMX_PSUBW : MMXI_binop_rm<0xF9, "psubw", sub, v4i16>; +defm MMX_PSUBD : MMXI_binop_rm<0xFA, "psubd", sub, v2i32>; +defm MMX_PSUBQ : MMXI_binop_rm<0xFB, "psubq", sub, v1i64>; + +defm MMX_PSUBSB : MMXI_binop_rm_int<0xE8, "psubsb" , int_x86_mmx_psubs_b>; +defm MMX_PSUBSW : MMXI_binop_rm_int<0xE9, "psubsw" , int_x86_mmx_psubs_w>; + +defm MMX_PSUBUSB : MMXI_binop_rm_int<0xD8, "psubusb", int_x86_mmx_psubus_b>; +defm MMX_PSUBUSW : MMXI_binop_rm_int<0xD9, "psubusw", int_x86_mmx_psubus_w>; + +// -- Multiplication +defm MMX_PMULLW : MMXI_binop_rm<0xD5, "pmullw", mul, v4i16, 1>; + +defm MMX_PMULHW : MMXI_binop_rm_int<0xE5, "pmulhw", int_x86_mmx_pmulh_w, 1>; +defm MMX_PMULHUW : MMXI_binop_rm_int<0xE4, "pmulhuw", int_x86_mmx_pmulhu_w, 1>; +defm MMX_PMULUDQ : MMXI_binop_rm_int<0xF4, "pmuludq", int_x86_mmx_pmulu_dq, 1>; + +// -- Miscellanea +defm MMX_PMADDWD : MMXI_binop_rm_int<0xF5, "pmaddwd", int_x86_mmx_pmadd_wd, 1>; + +defm MMX_PAVGB : MMXI_binop_rm_int<0xE0, "pavgb", int_x86_mmx_pavg_b, 1>; +defm MMX_PAVGW : MMXI_binop_rm_int<0xE3, "pavgw", int_x86_mmx_pavg_w, 1>; + +defm MMX_PMINUB : MMXI_binop_rm_int<0xDA, "pminub", int_x86_mmx_pminu_b, 1>; +defm MMX_PMINSW : MMXI_binop_rm_int<0xEA, "pminsw", int_x86_mmx_pmins_w, 1>; + +defm MMX_PMAXUB : MMXI_binop_rm_int<0xDE, "pmaxub", int_x86_mmx_pmaxu_b, 1>; +defm MMX_PMAXSW : MMXI_binop_rm_int<0xEE, "pmaxsw", int_x86_mmx_pmaxs_w, 1>; + +defm MMX_PSADBW : MMXI_binop_rm_int<0xE0, "psadbw", int_x86_mmx_psad_bw, 1>; + +// Logical Instructions +defm MMX_PAND : MMXI_binop_rm_v1i64<0xDB, "pand", and, 1>; +defm MMX_POR : MMXI_binop_rm_v1i64<0xEB, "por" , or, 1>; +defm MMX_PXOR : MMXI_binop_rm_v1i64<0xEF, "pxor", xor, 1>; + +let isTwoAddress = 1 in { + def MMX_PANDNrr : MMXI<0xDF, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "pandn\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, (v1i64 (and (vnot VR64:$src1), + VR64:$src2)))]>; + def MMX_PANDNrm : MMXI<0xDF, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "pandn\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, (v1i64 (and (vnot VR64:$src1), + (load addr:$src2))))]>; +} + +// Shift Instructions +defm MMX_PSRLW : MMXI_binop_rmi_int<0xD1, 0x71, MRM2r, "psrlw", + int_x86_mmx_psrl_w, int_x86_mmx_psrli_w>; +defm MMX_PSRLD : MMXI_binop_rmi_int<0xD2, 0x72, MRM2r, "psrld", + int_x86_mmx_psrl_d, int_x86_mmx_psrli_d>; +defm MMX_PSRLQ : MMXI_binop_rmi_int<0xD3, 0x73, MRM2r, "psrlq", + int_x86_mmx_psrl_q, int_x86_mmx_psrli_q>; + +defm MMX_PSLLW : MMXI_binop_rmi_int<0xF1, 0x71, MRM6r, "psllw", + int_x86_mmx_psll_w, int_x86_mmx_pslli_w>; +defm MMX_PSLLD : MMXI_binop_rmi_int<0xF2, 0x72, MRM6r, "pslld", + int_x86_mmx_psll_d, int_x86_mmx_pslli_d>; +defm MMX_PSLLQ : MMXI_binop_rmi_int<0xF3, 0x73, MRM6r, "psllq", + int_x86_mmx_psll_q, int_x86_mmx_pslli_q>; + +defm MMX_PSRAW : MMXI_binop_rmi_int<0xE1, 0x71, MRM4r, "psraw", + int_x86_mmx_psra_w, int_x86_mmx_psrai_w>; +defm MMX_PSRAD : MMXI_binop_rmi_int<0xE2, 0x72, MRM4r, "psrad", + int_x86_mmx_psra_d, int_x86_mmx_psrai_d>; + +// Comparison Instructions +defm MMX_PCMPEQB : MMXI_binop_rm_int<0x74, "pcmpeqb", int_x86_mmx_pcmpeq_b>; +defm MMX_PCMPEQW : MMXI_binop_rm_int<0x75, "pcmpeqw", int_x86_mmx_pcmpeq_w>; +defm MMX_PCMPEQD : MMXI_binop_rm_int<0x76, "pcmpeqd", int_x86_mmx_pcmpeq_d>; + +defm MMX_PCMPGTB : MMXI_binop_rm_int<0x64, "pcmpgtb", int_x86_mmx_pcmpgt_b>; +defm MMX_PCMPGTW : MMXI_binop_rm_int<0x65, "pcmpgtw", int_x86_mmx_pcmpgt_w>; +defm MMX_PCMPGTD : MMXI_binop_rm_int<0x66, "pcmpgtd", int_x86_mmx_pcmpgt_d>; + +// Conversion Instructions + +// -- Unpack Instructions +let isTwoAddress = 1 in { + // Unpack High Packed Data Instructions + def MMX_PUNPCKHBWrr : MMXI<0x68, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpckhbw\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v8i8 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKH_shuffle_mask)))]>; + def MMX_PUNPCKHBWrm : MMXI<0x68, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpckhbw\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v8i8 (vector_shuffle VR64:$src1, + (bc_v8i8 (load_mmx addr:$src2)), + MMX_UNPCKH_shuffle_mask)))]>; + + def MMX_PUNPCKHWDrr : MMXI<0x69, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpckhwd\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKH_shuffle_mask)))]>; + def MMX_PUNPCKHWDrm : MMXI<0x69, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpckhwd\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle VR64:$src1, + (bc_v4i16 (load_mmx addr:$src2)), + MMX_UNPCKH_shuffle_mask)))]>; + + def MMX_PUNPCKHDQrr : MMXI<0x6A, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpckhdq\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v2i32 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKH_shuffle_mask)))]>; + def MMX_PUNPCKHDQrm : MMXI<0x6A, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpckhdq\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v2i32 (vector_shuffle VR64:$src1, + (bc_v2i32 (load_mmx addr:$src2)), + MMX_UNPCKH_shuffle_mask)))]>; + + // Unpack Low Packed Data Instructions + def MMX_PUNPCKLBWrr : MMXI<0x60, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpcklbw\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v8i8 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKL_shuffle_mask)))]>; + def MMX_PUNPCKLBWrm : MMXI<0x60, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpcklbw\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v8i8 (vector_shuffle VR64:$src1, + (bc_v8i8 (load_mmx addr:$src2)), + MMX_UNPCKL_shuffle_mask)))]>; + + def MMX_PUNPCKLWDrr : MMXI<0x61, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpcklwd\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKL_shuffle_mask)))]>; + def MMX_PUNPCKLWDrm : MMXI<0x61, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpcklwd\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle VR64:$src1, + (bc_v4i16 (load_mmx addr:$src2)), + MMX_UNPCKL_shuffle_mask)))]>; + + def MMX_PUNPCKLDQrr : MMXI<0x62, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, VR64:$src2), + "punpckldq\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v2i32 (vector_shuffle VR64:$src1, VR64:$src2, + MMX_UNPCKL_shuffle_mask)))]>; + def MMX_PUNPCKLDQrm : MMXI<0x62, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2), + "punpckldq\t{$src2, $dst|$dst, $src2}", + [(set VR64:$dst, + (v2i32 (vector_shuffle VR64:$src1, + (bc_v2i32 (load_mmx addr:$src2)), + MMX_UNPCKL_shuffle_mask)))]>; +} + +// -- Pack Instructions +defm MMX_PACKSSWB : MMXI_binop_rm_int<0x63, "packsswb", int_x86_mmx_packsswb>; +defm MMX_PACKSSDW : MMXI_binop_rm_int<0x6B, "packssdw", int_x86_mmx_packssdw>; +defm MMX_PACKUSWB : MMXI_binop_rm_int<0x67, "packuswb", int_x86_mmx_packuswb>; + +// -- Shuffle Instructions +def MMX_PSHUFWri : MMXIi8<0x70, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, i8imm:$src2), + "pshufw\t{$src2, $src1, $dst|$dst, $src1, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle + VR64:$src1, (undef), + MMX_PSHUFW_shuffle_mask:$src2)))]>; +def MMX_PSHUFWmi : MMXIi8<0x70, MRMSrcMem, + (outs VR64:$dst), (ins i64mem:$src1, i8imm:$src2), + "pshufw\t{$src2, $src1, $dst|$dst, $src1, $src2}", + [(set VR64:$dst, + (v4i16 (vector_shuffle + (bc_v4i16 (load_mmx addr:$src1)), + (undef), + MMX_PSHUFW_shuffle_mask:$src2)))]>; + +// -- Conversion Instructions +let neverHasSideEffects = 1 in { +def MMX_CVTPD2PIrr : MMX2I<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src), + "cvtpd2pi\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTPD2PIrm : MMX2I<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src), + "cvtpd2pi\t{$src, $dst|$dst, $src}", []>; + +def MMX_CVTPI2PDrr : MMX2I<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src), + "cvtpi2pd\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTPI2PDrm : MMX2I<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), + "cvtpi2pd\t{$src, $dst|$dst, $src}", []>; + +def MMX_CVTPI2PSrr : MMXI<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src), + "cvtpi2ps\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTPI2PSrm : MMXI<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src), + "cvtpi2ps\t{$src, $dst|$dst, $src}", []>; + +def MMX_CVTPS2PIrr : MMXI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src), + "cvtps2pi\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTPS2PIrm : MMXI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src), + "cvtps2pi\t{$src, $dst|$dst, $src}", []>; + +def MMX_CVTTPD2PIrr : MMX2I<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src), + "cvttpd2pi\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTTPD2PIrm : MMX2I<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src), + "cvttpd2pi\t{$src, $dst|$dst, $src}", []>; + +def MMX_CVTTPS2PIrr : MMXI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src), + "cvttps2pi\t{$src, $dst|$dst, $src}", []>; +let mayLoad = 1 in +def MMX_CVTTPS2PIrm : MMXI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src), + "cvttps2pi\t{$src, $dst|$dst, $src}", []>; +} // end neverHasSideEffects + + +// Extract / Insert +def MMX_X86pextrw : SDNode<"X86ISD::PEXTRW", SDTypeProfile<1, 2, []>, []>; +def MMX_X86pinsrw : SDNode<"X86ISD::PINSRW", SDTypeProfile<1, 3, []>, []>; + +def MMX_PEXTRWri : MMXIi8<0xC5, MRMSrcReg, + (outs GR32:$dst), (ins VR64:$src1, i16i8imm:$src2), + "pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}", + [(set GR32:$dst, (MMX_X86pextrw (v4i16 VR64:$src1), + (iPTR imm:$src2)))]>; +let isTwoAddress = 1 in { + def MMX_PINSRWrri : MMXIi8<0xC4, MRMSrcReg, + (outs VR64:$dst), (ins VR64:$src1, GR32:$src2, i16i8imm:$src3), + "pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}", + [(set VR64:$dst, (v4i16 (MMX_X86pinsrw (v4i16 VR64:$src1), + GR32:$src2, (iPTR imm:$src3))))]>; + def MMX_PINSRWrmi : MMXIi8<0xC4, MRMSrcMem, + (outs VR64:$dst), (ins VR64:$src1, i16mem:$src2, i16i8imm:$src3), + "pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}", + [(set VR64:$dst, + (v4i16 (MMX_X86pinsrw (v4i16 VR64:$src1), + (i32 (anyext (loadi16 addr:$src2))), + (iPTR imm:$src3))))]>; +} + +// Mask creation +def MMX_PMOVMSKBrr : MMXI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR64:$src), + "pmovmskb\t{$src, $dst|$dst, $src}", + [(set GR32:$dst, (int_x86_mmx_pmovmskb VR64:$src))]>; // Misc. -def MOVNTQ : I<0xE7, MRMDestMem, (ops i64mem:$dst, VR64:$src), - "movntq {$src, $dst|$dst, $src}", []>, TB, - Requires<[HasMMX]>; +let Uses = [EDI] in +def MMX_MASKMOVQ : MMXI<0xF7, MRMDestMem, (outs), (ins VR64:$src, VR64:$mask), + "maskmovq\t{$mask, $src|$src, $mask}", + [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, EDI)]>; + +//===----------------------------------------------------------------------===// +// Alias Instructions +//===----------------------------------------------------------------------===// + +// Alias instructions that map zero vector to pxor. +let isReMaterializable = 1 in { + def MMX_V_SET0 : MMXI<0xEF, MRMInitReg, (outs VR64:$dst), (ins), + "pxor\t$dst, $dst", + [(set VR64:$dst, (v2i32 immAllZerosV))]>; + def MMX_V_SETALLONES : MMXI<0x76, MRMInitReg, (outs VR64:$dst), (ins), + "pcmpeqd\t$dst, $dst", + [(set VR64:$dst, (v2i32 immAllOnesV))]>; +} -def MASKMOVQ : I<0xF7, MRMDestMem, (ops VR64:$src, VR64:$mask), - "maskmovq {$mask, $src|$src, $mask}", []>, TB, - Requires<[HasMMX]>; +let Predicates = [HasMMX] in { + def : Pat<(v1i64 immAllZerosV), (MMX_V_SET0)>; + def : Pat<(v4i16 immAllZerosV), (MMX_V_SET0)>; + def : Pat<(v8i8 immAllZerosV), (MMX_V_SET0)>; +} //===----------------------------------------------------------------------===// // Non-Instruction Patterns @@ -118,14 +521,133 @@ def MASKMOVQ : I<0xF7, MRMDestMem, (ops VR64:$src, VR64:$mask), // Store 64-bit integer vector values. def : Pat<(store (v8i8 VR64:$src), addr:$dst), - (MOVQ64mr addr:$dst, VR64:$src)>; + (MMX_MOVQ64mr addr:$dst, VR64:$src)>; def : Pat<(store (v4i16 VR64:$src), addr:$dst), - (MOVQ64mr addr:$dst, VR64:$src)>; + (MMX_MOVQ64mr addr:$dst, VR64:$src)>; +def : Pat<(store (v2i32 VR64:$src), addr:$dst), + (MMX_MOVQ64mr addr:$dst, VR64:$src)>; +def : Pat<(store (v1i64 VR64:$src), addr:$dst), + (MMX_MOVQ64mr addr:$dst, VR64:$src)>; // Bit convert. +def : Pat<(v8i8 (bitconvert (v1i64 VR64:$src))), (v8i8 VR64:$src)>; def : Pat<(v8i8 (bitconvert (v2i32 VR64:$src))), (v8i8 VR64:$src)>; def : Pat<(v8i8 (bitconvert (v4i16 VR64:$src))), (v8i8 VR64:$src)>; +def : Pat<(v4i16 (bitconvert (v1i64 VR64:$src))), (v4i16 VR64:$src)>; def : Pat<(v4i16 (bitconvert (v2i32 VR64:$src))), (v4i16 VR64:$src)>; def : Pat<(v4i16 (bitconvert (v8i8 VR64:$src))), (v4i16 VR64:$src)>; +def : Pat<(v2i32 (bitconvert (v1i64 VR64:$src))), (v2i32 VR64:$src)>; def : Pat<(v2i32 (bitconvert (v4i16 VR64:$src))), (v2i32 VR64:$src)>; def : Pat<(v2i32 (bitconvert (v8i8 VR64:$src))), (v2i32 VR64:$src)>; +def : Pat<(v1i64 (bitconvert (v2i32 VR64:$src))), (v1i64 VR64:$src)>; +def : Pat<(v1i64 (bitconvert (v4i16 VR64:$src))), (v1i64 VR64:$src)>; +def : Pat<(v1i64 (bitconvert (v8i8 VR64:$src))), (v1i64 VR64:$src)>; + +// 64-bit bit convert. +def : Pat<(v1i64 (bitconvert (i64 GR64:$src))), + (MMX_MOVD64to64rr GR64:$src)>; +def : Pat<(v2i32 (bitconvert (i64 GR64:$src))), + (MMX_MOVD64to64rr GR64:$src)>; +def : Pat<(v4i16 (bitconvert (i64 GR64:$src))), + (MMX_MOVD64to64rr GR64:$src)>; +def : Pat<(v8i8 (bitconvert (i64 GR64:$src))), + (MMX_MOVD64to64rr GR64:$src)>; +def : Pat<(i64 (bitconvert (v1i64 VR64:$src))), + (MMX_MOVD64from64rr VR64:$src)>; +def : Pat<(i64 (bitconvert (v2i32 VR64:$src))), + (MMX_MOVD64from64rr VR64:$src)>; +def : Pat<(i64 (bitconvert (v4i16 VR64:$src))), + (MMX_MOVD64from64rr VR64:$src)>; +def : Pat<(i64 (bitconvert (v8i8 VR64:$src))), + (MMX_MOVD64from64rr VR64:$src)>; + +// Move scalar to XMM zero-extended +// movd to XMM register zero-extends +let AddedComplexity = 15 in { + def : Pat<(v8i8 (vector_shuffle immAllZerosV_bc, + (bc_v8i8 (v2i32 (scalar_to_vector GR32:$src))), + MMX_MOVL_shuffle_mask)), + (MMX_MOVZDI2PDIrr GR32:$src)>; + def : Pat<(v4i16 (vector_shuffle immAllZerosV_bc, + (bc_v4i16 (v2i32 (scalar_to_vector GR32:$src))), + MMX_MOVL_shuffle_mask)), + (MMX_MOVZDI2PDIrr GR32:$src)>; +} + +// Scalar to v4i16 / v8i8. The source may be a GR32, but only the lower +// 8 or 16-bits matter. +def : Pat<(bc_v8i8 (v2i32 (scalar_to_vector GR32:$src))), + (MMX_MOVD64rr GR32:$src)>; +def : Pat<(bc_v4i16 (v2i32 (scalar_to_vector GR32:$src))), + (MMX_MOVD64rr GR32:$src)>; + +// Patterns to perform canonical versions of vector shuffling. +let AddedComplexity = 10 in { + def : Pat<(v8i8 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKL_v_undef_shuffle_mask)), + (MMX_PUNPCKLBWrr VR64:$src, VR64:$src)>; + def : Pat<(v4i16 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKL_v_undef_shuffle_mask)), + (MMX_PUNPCKLWDrr VR64:$src, VR64:$src)>; + def : Pat<(v2i32 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKL_v_undef_shuffle_mask)), + (MMX_PUNPCKLDQrr VR64:$src, VR64:$src)>; +} + +let AddedComplexity = 10 in { + def : Pat<(v8i8 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKH_v_undef_shuffle_mask)), + (MMX_PUNPCKHBWrr VR64:$src, VR64:$src)>; + def : Pat<(v4i16 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKH_v_undef_shuffle_mask)), + (MMX_PUNPCKHWDrr VR64:$src, VR64:$src)>; + def : Pat<(v2i32 (vector_shuffle VR64:$src, (undef), + MMX_UNPCKH_v_undef_shuffle_mask)), + (MMX_PUNPCKHDQrr VR64:$src, VR64:$src)>; +} + +// Patterns to perform vector shuffling with a zeroed out vector. +let AddedComplexity = 20 in { + def : Pat<(bc_v2i32 (vector_shuffle immAllZerosV, + (v2i32 (scalar_to_vector (load_mmx addr:$src))), + MMX_UNPCKL_shuffle_mask)), + (MMX_PUNPCKLDQrm VR64:$src, VR64:$src)>; +} + +// Some special case PANDN patterns. +// FIXME: Get rid of these. +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v2i32 immAllOnesV))), + VR64:$src2)), + (MMX_PANDNrr VR64:$src1, VR64:$src2)>; +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV_bc))), + VR64:$src2)), + (MMX_PANDNrr VR64:$src1, VR64:$src2)>; +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV_bc))), + VR64:$src2)), + (MMX_PANDNrr VR64:$src1, VR64:$src2)>; + +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v2i32 immAllOnesV))), + (load addr:$src2))), + (MMX_PANDNrm VR64:$src1, addr:$src2)>; +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV_bc))), + (load addr:$src2))), + (MMX_PANDNrm VR64:$src1, addr:$src2)>; +def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV_bc))), + (load addr:$src2))), + (MMX_PANDNrm VR64:$src1, addr:$src2)>; + +// Move MMX to lower 64-bit of XMM +def : Pat<(v2i64 (scalar_to_vector (i64 (bitconvert VR64:$src)))), + (v2i64 (MMX_MOVQ2DQrr VR64:$src))>; + +// Move lower 64-bit of XMM to MMX. +def : Pat<(v2i32 (bitconvert (i64 (vector_extract (v2i64 VR128:$src), + (iPTR 0))))), + (v2i32 (MMX_MOVDQ2Qrr VR128:$src))>; +def : Pat<(v4i16 (bitconvert (i64 (vector_extract (v2i64 VR128:$src), + (iPTR 0))))), + (v4i16 (MMX_MOVDQ2Qrr VR128:$src))>; +def : Pat<(v8i8 (bitconvert (i64 (vector_extract (v2i64 VR128:$src), + (iPTR 0))))), + (v8i8 (MMX_MOVDQ2Qrr VR128:$src))>; +