X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTarget%2FPowerPC%2FPPCInstrAltivec.td;h=2de852b48434748a8127b893e7355aadf156576a;hp=3174892c1137c5f0713129accb35e59b69b2f450;hb=b69d556c370b32dee9f64d8250e51aad33963cc2;hpb=b5c4d17a6c03af1ec8e3fd4fd9d5b07b9c2694a4 diff --git a/lib/Target/PowerPC/PPCInstrAltivec.td b/lib/Target/PowerPC/PPCInstrAltivec.td index 3174892c113..2de852b4843 100644 --- a/lib/Target/PowerPC/PPCInstrAltivec.td +++ b/lib/Target/PowerPC/PPCInstrAltivec.td @@ -1,10 +1,10 @@ -//===- PPCInstrAltivec.td - The PowerPC Altivec Extension --*- tablegen -*-===// -// +//===-- PPCInstrAltivec.td - The PowerPC Altivec Extension -*- tablegen -*-===// +// // The LLVM Compiler Infrastructure // -// This file was developed by Chris Lattner 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. +// //===----------------------------------------------------------------------===// // // This file describes the Altivec extension to the PowerPC instruction set. @@ -15,391 +15,717 @@ // Altivec transformation functions and pattern fragments. // -// VSPLT_get_imm xform function: convert vector_shuffle mask to VSPLT* imm. -def VSPLT_get_imm : SDNodeXForm; + +def vpkuhum_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUHUMShuffleMask(cast(N), 0, *CurDAG); +}]>; +def vpkuwum_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUWUMShuffleMask(cast(N), 0, *CurDAG); +}]>; +def vpkuhum_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUHUMShuffleMask(cast(N), 1, *CurDAG); +}]>; +def vpkuwum_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUWUMShuffleMask(cast(N), 1, *CurDAG); }]>; -def VSPLT_shuffle_mask : PatLeaf<(build_vector), [{ - return PPC::isSplatShuffleMask(N); -}], VSPLT_get_imm>; +// These fragments are provided for little-endian, where the inputs must be +// swapped for correct semantics. +def vpkuhum_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUHUMShuffleMask(cast(N), 2, *CurDAG); +}]>; +def vpkuwum_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVPKUWUMShuffleMask(cast(N), 2, *CurDAG); +}]>; + +def vmrglb_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 1, 0, *CurDAG); +}]>; +def vmrglh_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 2, 0, *CurDAG); +}]>; +def vmrglw_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 4, 0, *CurDAG); +}]>; +def vmrghb_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 1, 0, *CurDAG); +}]>; +def vmrghh_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 2, 0, *CurDAG); +}]>; +def vmrghw_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 4, 0, *CurDAG); +}]>; + + +def vmrglb_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 1, 1, *CurDAG); +}]>; +def vmrglh_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 2, 1, *CurDAG); +}]>; +def vmrglw_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 4, 1, *CurDAG); +}]>; +def vmrghb_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 1, 1, *CurDAG); +}]>; +def vmrghh_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 2, 1, *CurDAG); +}]>; +def vmrghw_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 4, 1, *CurDAG); +}]>; + + +// These fragments are provided for little-endian, where the inputs must be +// swapped for correct semantics. +def vmrglb_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 1, 2, *CurDAG); +}]>; +def vmrglh_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 2, 2, *CurDAG); +}]>; +def vmrglw_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGLShuffleMask(cast(N), 4, 2, *CurDAG); +}]>; +def vmrghb_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 1, 2, *CurDAG); +}]>; +def vmrghh_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 2, 2, *CurDAG); +}]>; +def vmrghw_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVMRGHShuffleMask(cast(N), 4, 2, *CurDAG); +}]>; + + +def VSLDOI_get_imm : SDNodeXForm; +def vsldoi_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVSLDOIShuffleMask(N, 0, *CurDAG) != -1; +}], VSLDOI_get_imm>; + + +/// VSLDOI_unary* - These are used to match vsldoi(X,X), which is turned into +/// vector_shuffle(X,undef,mask) by the dag combiner. +def VSLDOI_unary_get_imm : SDNodeXForm; +def vsldoi_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVSLDOIShuffleMask(N, 1, *CurDAG) != -1; +}], VSLDOI_unary_get_imm>; + + +/// VSLDOI_swapped* - These fragments are provided for little-endian, where +/// the inputs must be swapped for correct semantics. +def VSLDOI_swapped_get_imm : SDNodeXForm; +def vsldoi_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isVSLDOIShuffleMask(N, 2, *CurDAG) != -1; +}], VSLDOI_get_imm>; + + +// VSPLT*_get_imm xform function: convert vector_shuffle mask to VSPLT* imm. +def VSPLTB_get_imm : SDNodeXForm; +def vspltb_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isSplatShuffleMask(cast(N), 1); +}], VSPLTB_get_imm>; +def VSPLTH_get_imm : SDNodeXForm; +def vsplth_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isSplatShuffleMask(cast(N), 2); +}], VSPLTH_get_imm>; +def VSPLTW_get_imm : SDNodeXForm; +def vspltw_shuffle : PatFrag<(ops node:$lhs, node:$rhs), + (vector_shuffle node:$lhs, node:$rhs), [{ + return PPC::isSplatShuffleMask(cast(N), 4); +}], VSPLTW_get_imm>; // VSPLTISB_get_imm xform function: convert build_vector to VSPLTISB imm. def VSPLTISB_get_imm : SDNodeXForm; def vecspltisb : PatLeaf<(build_vector), [{ - return PPC::isVecSplatImm(N, 1); + return PPC::get_VSPLTI_elt(N, 1, *CurDAG).getNode() != 0; }], VSPLTISB_get_imm>; // VSPLTISH_get_imm xform function: convert build_vector to VSPLTISH imm. def VSPLTISH_get_imm : SDNodeXForm; def vecspltish : PatLeaf<(build_vector), [{ - return PPC::isVecSplatImm(N, 2); + return PPC::get_VSPLTI_elt(N, 2, *CurDAG).getNode() != 0; }], VSPLTISH_get_imm>; // VSPLTISW_get_imm xform function: convert build_vector to VSPLTISW imm. def VSPLTISW_get_imm : SDNodeXForm; def vecspltisw : PatLeaf<(build_vector), [{ - return PPC::isVecSplatImm(N, 4); + return PPC::get_VSPLTI_elt(N, 4, *CurDAG).getNode() != 0; }], VSPLTISW_get_imm>; -class isVDOT { // vector dot instruction. - list Defs = [CR6]; - bit RC = 1; -} - //===----------------------------------------------------------------------===// // Helpers for defining instructions that directly correspond to intrinsics. -// VA1a_Int - A VAForm_1a intrinsic definition. -class VA1a_Int xo, string opc, Intrinsic IntID> - : VAForm_1a; +// VA1a_Int_Ty - A VAForm_1a intrinsic definition of specific type. +class VA1a_Int_Ty xo, string opc, Intrinsic IntID, ValueType Ty> + : VAForm_1a; + +// VA1a_Int_Ty2 - A VAForm_1a intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VA1a_Int_Ty2 xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VAForm_1a; + +// VA1a_Int_Ty3 - A VAForm_1a intrinsic definition where there are two +// input types and an output type. +class VA1a_Int_Ty3 xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType In1Ty, ValueType In2Ty> + : VAForm_1a; -// VX1_Int - A VXForm_1 intrinsic definition. -class VX1_Int xo, string asmstr, Intrinsic IntID> - : VXForm_1; +// VX1_Int_Ty - A VXForm_1 intrinsic definition of specific type. +class VX1_Int_Ty xo, string opc, Intrinsic IntID, ValueType Ty> + : VXForm_1; -// VX2_Int - A VXForm_2 intrinsic definition. -class VX2_Int xo, string asmstr, Intrinsic IntID> - : VXForm_2; +// VX1_Int_Ty2 - A VXForm_1 intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VX1_Int_Ty2 xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VXForm_1; + +// VX1_Int_Ty3 - A VXForm_1 intrinsic definition where there are two +// input types and an output type. +class VX1_Int_Ty3 xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType In1Ty, ValueType In2Ty> + : VXForm_1; + +// VX2_Int_SP - A VXForm_2 intrinsic definition of vector single-precision type. +class VX2_Int_SP xo, string opc, Intrinsic IntID> + : VXForm_2; + +// VX2_Int_Ty2 - A VXForm_2 intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VX2_Int_Ty2 xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VXForm_2; + +class VXBX_Int_Ty xo, string opc, Intrinsic IntID, ValueType Ty> + : VXForm_BX; + +class VXCR_Int_Ty xo, string opc, Intrinsic IntID, ValueType Ty> + : VXForm_CR; //===----------------------------------------------------------------------===// // Instruction Definitions. -def IMPLICIT_DEF_VRRC : Pseudo<(ops VRRC:$rD), "; $rD = IMPLICIT_DEF_VRRC", - [(set VRRC:$rD, (v4f32 (undef)))]>; - -let isLoad = 1, PPC970_Unit = 2 in { // Loads. -def LVEBX: XForm_1<31, 7, (ops VRRC:$vD, memrr:$src), - "lvebx $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>; -def LVEHX: XForm_1<31, 39, (ops VRRC:$vD, memrr:$src), - "lvehx $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>; -def LVEWX: XForm_1<31, 71, (ops VRRC:$vD, memrr:$src), - "lvewx $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>; -def LVX : XForm_1<31, 103, (ops VRRC:$vD, memrr:$src), - "lvx $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>; -def LVXL : XForm_1<31, 359, (ops VRRC:$vD, memrr:$src), - "lvxl $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>; +def HasAltivec : Predicate<"PPCSubTarget->hasAltivec()">; +let Predicates = [HasAltivec] in { + +def DSS : DSS_Form<0, 822, (outs), (ins u5imm:$STRM), + "dss $STRM", IIC_LdStLoad /*FIXME*/, [(int_ppc_altivec_dss imm:$STRM)]>, + Deprecated { + let A = 0; + let B = 0; +} + +def DSSALL : DSS_Form<1, 822, (outs), (ins), + "dssall", IIC_LdStLoad /*FIXME*/, [(int_ppc_altivec_dssall)]>, + Deprecated { + let STRM = 0; + let A = 0; + let B = 0; } -def LVSL : XForm_1<31, 6, (ops VRRC:$vD, memrr:$src), - "lvsl $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>, +def DST : DSS_Form<0, 342, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB), + "dst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dst i32:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + +def DSTT : DSS_Form<1, 342, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB), + "dstt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dstt i32:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + +def DSTST : DSS_Form<0, 374, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB), + "dstst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dstst i32:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + +def DSTSTT : DSS_Form<1, 374, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB), + "dststt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dststt i32:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + +let isCodeGenOnly = 1 in { + // The very same instructions as above, but formally matching 64bit registers. + def DST64 : DSS_Form<0, 342, (outs), (ins u5imm:$STRM, g8rc:$rA, gprc:$rB), + "dst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dst i64:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + + def DSTT64 : DSS_Form<1, 342, (outs), (ins u5imm:$STRM, g8rc:$rA, gprc:$rB), + "dstt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dstt i64:$rA, i32:$rB, imm:$STRM)]>, + Deprecated; + + def DSTST64 : DSS_Form<0, 374, (outs), (ins u5imm:$STRM, g8rc:$rA, gprc:$rB), + "dstst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dstst i64:$rA, i32:$rB, + imm:$STRM)]>, + Deprecated; + + def DSTSTT64 : DSS_Form<1, 374, (outs), (ins u5imm:$STRM, g8rc:$rA, gprc:$rB), + "dststt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, + [(int_ppc_altivec_dststt i64:$rA, i32:$rB, + imm:$STRM)]>, + Deprecated; +} + +def MFVSCR : VXForm_4<1540, (outs vrrc:$vD), (ins), + "mfvscr $vD", IIC_LdStStore, + [(set v8i16:$vD, (int_ppc_altivec_mfvscr))]>; +def MTVSCR : VXForm_5<1604, (outs), (ins vrrc:$vB), + "mtvscr $vB", IIC_LdStLoad, + [(int_ppc_altivec_mtvscr v4i32:$vB)]>; + +let canFoldAsLoad = 1, PPC970_Unit = 2 in { // Loads. +def LVEBX: XForm_1<31, 7, (outs vrrc:$vD), (ins memrr:$src), + "lvebx $vD, $src", IIC_LdStLoad, + [(set v16i8:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>; +def LVEHX: XForm_1<31, 39, (outs vrrc:$vD), (ins memrr:$src), + "lvehx $vD, $src", IIC_LdStLoad, + [(set v8i16:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>; +def LVEWX: XForm_1<31, 71, (outs vrrc:$vD), (ins memrr:$src), + "lvewx $vD, $src", IIC_LdStLoad, + [(set v4i32:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>; +def LVX : XForm_1<31, 103, (outs vrrc:$vD), (ins memrr:$src), + "lvx $vD, $src", IIC_LdStLoad, + [(set v4i32:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>; +def LVXL : XForm_1<31, 359, (outs vrrc:$vD), (ins memrr:$src), + "lvxl $vD, $src", IIC_LdStLoad, + [(set v4i32:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>; +} + +def LVSL : XForm_1<31, 6, (outs vrrc:$vD), (ins memrr:$src), + "lvsl $vD, $src", IIC_LdStLoad, + [(set v16i8:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>, PPC970_Unit_LSU; -def LVSR : XForm_1<31, 38, (ops VRRC:$vD, memrr:$src), - "lvsl $vD, $src", LdStGeneral, - [(set VRRC:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>, +def LVSR : XForm_1<31, 38, (outs vrrc:$vD), (ins memrr:$src), + "lvsr $vD, $src", IIC_LdStLoad, + [(set v16i8:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>, PPC970_Unit_LSU; -let isStore = 1, noResults = 1, PPC970_Unit = 2 in { // Stores. -def STVEBX: XForm_8<31, 135, (ops VRRC:$rS, memrr:$dst), - "stvebx $rS, $dst", LdStGeneral, - [(int_ppc_altivec_stvebx VRRC:$rS, xoaddr:$dst)]>; -def STVEHX: XForm_8<31, 167, (ops VRRC:$rS, memrr:$dst), - "stvehx $rS, $dst", LdStGeneral, - [(int_ppc_altivec_stvehx VRRC:$rS, xoaddr:$dst)]>; -def STVEWX: XForm_8<31, 199, (ops VRRC:$rS, memrr:$dst), - "stvewx $rS, $dst", LdStGeneral, - [(int_ppc_altivec_stvewx VRRC:$rS, xoaddr:$dst)]>; -def STVX : XForm_8<31, 231, (ops VRRC:$rS, memrr:$dst), - "stvx $rS, $dst", LdStGeneral, - [(int_ppc_altivec_stvx VRRC:$rS, xoaddr:$dst)]>; -def STVXL : XForm_8<31, 487, (ops VRRC:$rS, memrr:$dst), - "stvxl $rS, $dst", LdStGeneral, - [(int_ppc_altivec_stvxl VRRC:$rS, xoaddr:$dst)]>; +let PPC970_Unit = 2 in { // Stores. +def STVEBX: XForm_8<31, 135, (outs), (ins vrrc:$rS, memrr:$dst), + "stvebx $rS, $dst", IIC_LdStStore, + [(int_ppc_altivec_stvebx v16i8:$rS, xoaddr:$dst)]>; +def STVEHX: XForm_8<31, 167, (outs), (ins vrrc:$rS, memrr:$dst), + "stvehx $rS, $dst", IIC_LdStStore, + [(int_ppc_altivec_stvehx v8i16:$rS, xoaddr:$dst)]>; +def STVEWX: XForm_8<31, 199, (outs), (ins vrrc:$rS, memrr:$dst), + "stvewx $rS, $dst", IIC_LdStStore, + [(int_ppc_altivec_stvewx v4i32:$rS, xoaddr:$dst)]>; +def STVX : XForm_8<31, 231, (outs), (ins vrrc:$rS, memrr:$dst), + "stvx $rS, $dst", IIC_LdStStore, + [(int_ppc_altivec_stvx v4i32:$rS, xoaddr:$dst)]>; +def STVXL : XForm_8<31, 487, (outs), (ins vrrc:$rS, memrr:$dst), + "stvxl $rS, $dst", IIC_LdStStore, + [(int_ppc_altivec_stvxl v4i32:$rS, xoaddr:$dst)]>; } let PPC970_Unit = 5 in { // VALU Operations. // VA-Form instructions. 3-input AltiVec ops. -def VMADDFP : VAForm_1<46, (ops VRRC:$vD, VRRC:$vA, VRRC:$vC, VRRC:$vB), - "vmaddfp $vD, $vA, $vC, $vB", VecFP, - [(set VRRC:$vD, (fadd (fmul VRRC:$vA, VRRC:$vC), - VRRC:$vB))]>, - Requires<[FPContractions]>; -def VNMSUBFP: VAForm_1<47, (ops VRRC:$vD, VRRC:$vA, VRRC:$vC, VRRC:$vB), - "vnmsubfp $vD, $vA, $vC, $vB", VecFP, - [(set VRRC:$vD, (fneg (fsub (fmul VRRC:$vA, VRRC:$vC), - VRRC:$vB)))]>, - Requires<[FPContractions]>; -def VMHADDSHS : VA1a_Int<32, "vmhaddshs", int_ppc_altivec_vmhaddshs>; -def VMHRADDSHS : VA1a_Int<33, "vmhraddshs", int_ppc_altivec_vmhraddshs>; -def VPERM : VA1a_Int<43, "vperm", int_ppc_altivec_vperm>; -def VSEL : VA1a_Int<42, "vsel", int_ppc_altivec_vsel>; - -def VSLDOI : VAForm_2<44, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB, u5imm:$SH), - "vsldoi $vD, $vA, $vB, $SH", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vsldoi VRRC:$vA, VRRC:$vB, - imm:$SH))]>; +let isCommutable = 1 in { +def VMADDFP : VAForm_1<46, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vC, vrrc:$vB), + "vmaddfp $vD, $vA, $vC, $vB", IIC_VecFP, + [(set v4f32:$vD, + (fma v4f32:$vA, v4f32:$vC, v4f32:$vB))]>; + +// FIXME: The fma+fneg pattern won't match because fneg is not legal. +def VNMSUBFP: VAForm_1<47, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vC, vrrc:$vB), + "vnmsubfp $vD, $vA, $vC, $vB", IIC_VecFP, + [(set v4f32:$vD, (fneg (fma v4f32:$vA, v4f32:$vC, + (fneg v4f32:$vB))))]>; + +def VMHADDSHS : VA1a_Int_Ty<32, "vmhaddshs", int_ppc_altivec_vmhaddshs, v8i16>; +def VMHRADDSHS : VA1a_Int_Ty<33, "vmhraddshs", int_ppc_altivec_vmhraddshs, + v8i16>; +def VMLADDUHM : VA1a_Int_Ty<34, "vmladduhm", int_ppc_altivec_vmladduhm, v8i16>; +} // isCommutable + +def VPERM : VA1a_Int_Ty3<43, "vperm", int_ppc_altivec_vperm, + v4i32, v4i32, v16i8>; +def VSEL : VA1a_Int_Ty<42, "vsel", int_ppc_altivec_vsel, v4i32>; + +// Shuffles. +def VSLDOI : VAForm_2<44, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB, u5imm:$SH), + "vsldoi $vD, $vA, $vB, $SH", IIC_VecFP, + [(set v16i8:$vD, + (vsldoi_shuffle:$SH v16i8:$vA, v16i8:$vB))]>; // VX-Form instructions. AltiVec arithmetic ops. -def VADDCUW : VXForm_1<384, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddcuw $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vaddcuw VRRC:$vA, VRRC:$vB))]>; -def VADDFP : VXForm_1<10, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddfp $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (fadd VRRC:$vA, VRRC:$vB))]>; +let isCommutable = 1 in { +def VADDFP : VXForm_1<10, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vaddfp $vD, $vA, $vB", IIC_VecFP, + [(set v4f32:$vD, (fadd v4f32:$vA, v4f32:$vB))]>; -def VADDUBM : VXForm_1<0, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddubm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VADDUHM : VXForm_1<64, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vadduhm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v8i16 VRRC:$vA), VRRC:$vB))]>; -def VADDUWM : VXForm_1<128, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vadduwm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v4i32 VRRC:$vA), VRRC:$vB))]>; +def VADDUBM : VXForm_1<0, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vaddubm $vD, $vA, $vB", IIC_VecGeneral, + [(set v16i8:$vD, (add v16i8:$vA, v16i8:$vB))]>; +def VADDUHM : VXForm_1<64, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vadduhm $vD, $vA, $vB", IIC_VecGeneral, + [(set v8i16:$vD, (add v8i16:$vA, v8i16:$vB))]>; +def VADDUWM : VXForm_1<128, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vadduwm $vD, $vA, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (add v4i32:$vA, v4i32:$vB))]>; -def VADDSBS : VXForm_1<768, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddsbs $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vaddsbs VRRC:$vA, VRRC:$vB))]>; -def VADDSHS : VXForm_1<832, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddshs $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vaddshs VRRC:$vA, VRRC:$vB))]>; -def VADDSWS : VXForm_1<896, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddsws $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vaddsws VRRC:$vA, VRRC:$vB))]>; - -def VADDUBS : VXForm_1<512, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vaddubs $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vaddubs VRRC:$vA, VRRC:$vB))]>; -def VADDUHS : VXForm_1<576, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vadduhs $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vadduhs VRRC:$vA, VRRC:$vB))]>; -def VADDUWS : VXForm_1<640, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vadduws $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vadduws VRRC:$vA, VRRC:$vB))]>; -def VAND : VXForm_1<1028, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vand $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (and (v4i32 VRRC:$vA), VRRC:$vB))]>; -def VANDC : VXForm_1<1092, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vandc $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (and (v4i32 VRRC:$vA), (vnot VRRC:$vB)))]>; - -def VCFSX : VXForm_1<842, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vcfsx $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfsx VRRC:$vB, imm:$UIMM))]>; -def VCFUX : VXForm_1<778, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vcfux $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfux VRRC:$vB, imm:$UIMM))]>; -def VCTSXS : VXForm_1<970, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vctsxs $vD, $vB, $UIMM", VecFP, - []>; -def VCTUXS : VXForm_1<906, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vctuxs $vD, $vB, $UIMM", VecFP, - []>; -def VEXPTEFP : VXForm_2<394, (ops VRRC:$vD, VRRC:$vB), - "vexptefp $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vexptefp VRRC:$vB))]>; -def VLOGEFP : VXForm_2<458, (ops VRRC:$vD, VRRC:$vB), - "vlogefp $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vlogefp VRRC:$vB))]>; -def VMAXFP : VXForm_1<1034, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vmaxfp $vD, $vA, $vB", VecFP, - []>; -def VMINFP : VXForm_1<1098, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vminfp $vD, $vA, $vB", VecFP, - []>; - -def VMRGHH : VX1_Int<76 , "vmrghh $vD, $vA, $vB", int_ppc_altivec_vmrghh>; -def VMRGHW : VX1_Int<140, "vmrghw $vD, $vA, $vB", int_ppc_altivec_vmrghw>; -def VMRGLH : VX1_Int<332, "vmrglh $vD, $vA, $vB", int_ppc_altivec_vmrglh>; -def VMRGLW : VX1_Int<396, "vmrglw $vD, $vA, $vB", int_ppc_altivec_vmrglw>; - -def VMSUMMBM : VA1a_Int<37, "vmsummbm", int_ppc_altivec_vmsummbm>; -def VMSUMSHM : VA1a_Int<40, "vmsumshm", int_ppc_altivec_vmsumshm>; -def VMSUMSHS : VA1a_Int<41, "vmsumshs", int_ppc_altivec_vmsumshs>; -def VMSUMUBM : VA1a_Int<36, "vmsumubm", int_ppc_altivec_vmsumubm>; -def VMSUMUHM : VA1a_Int<38, "vmsumuhm", int_ppc_altivec_vmsumuhm>; -def VMSUMUHS : VA1a_Int<39, "vmsumuhs", int_ppc_altivec_vmsumuhs>; - -def VMULESB : VX1_Int<776, "vmulesb $vD, $vA, $vB", int_ppc_altivec_vmulesb>; -def VMULESH : VX1_Int<840, "vmulesh $vD, $vA, $vB", int_ppc_altivec_vmulesh>; -def VMULEUB : VX1_Int<520, "vmuleub $vD, $vA, $vB", int_ppc_altivec_vmuleub>; -def VMULEUH : VX1_Int<584, "vmuleuh $vD, $vA, $vB", int_ppc_altivec_vmuleuh>; -def VMULOSB : VX1_Int<264, "vmulosb $vD, $vA, $vB", int_ppc_altivec_vmulosb>; -def VMULOSH : VX1_Int<328, "vmulosh $vD, $vA, $vB", int_ppc_altivec_vmulosh>; -def VMULOUB : VX1_Int< 8, "vmuloub $vD, $vA, $vB", int_ppc_altivec_vmuloub>; -def VMULOUH : VX1_Int< 72, "vmulouh $vD, $vA, $vB", int_ppc_altivec_vmulouh>; +def VADDCUW : VX1_Int_Ty<384, "vaddcuw", int_ppc_altivec_vaddcuw, v4i32>; +def VADDSBS : VX1_Int_Ty<768, "vaddsbs", int_ppc_altivec_vaddsbs, v16i8>; +def VADDSHS : VX1_Int_Ty<832, "vaddshs", int_ppc_altivec_vaddshs, v8i16>; +def VADDSWS : VX1_Int_Ty<896, "vaddsws", int_ppc_altivec_vaddsws, v4i32>; +def VADDUBS : VX1_Int_Ty<512, "vaddubs", int_ppc_altivec_vaddubs, v16i8>; +def VADDUHS : VX1_Int_Ty<576, "vadduhs", int_ppc_altivec_vadduhs, v8i16>; +def VADDUWS : VX1_Int_Ty<640, "vadduws", int_ppc_altivec_vadduws, v4i32>; +} // isCommutable + +let isCommutable = 1 in +def VAND : VXForm_1<1028, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vand $vD, $vA, $vB", IIC_VecFP, + [(set v4i32:$vD, (and v4i32:$vA, v4i32:$vB))]>; +def VANDC : VXForm_1<1092, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vandc $vD, $vA, $vB", IIC_VecFP, + [(set v4i32:$vD, (and v4i32:$vA, + (vnot_ppc v4i32:$vB)))]>; + +def VCFSX : VXForm_1<842, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vcfsx $vD, $vB, $UIMM", IIC_VecFP, + [(set v4f32:$vD, + (int_ppc_altivec_vcfsx v4i32:$vB, imm:$UIMM))]>; +def VCFUX : VXForm_1<778, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vcfux $vD, $vB, $UIMM", IIC_VecFP, + [(set v4f32:$vD, + (int_ppc_altivec_vcfux v4i32:$vB, imm:$UIMM))]>; +def VCTSXS : VXForm_1<970, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vctsxs $vD, $vB, $UIMM", IIC_VecFP, + [(set v4i32:$vD, + (int_ppc_altivec_vctsxs v4f32:$vB, imm:$UIMM))]>; +def VCTUXS : VXForm_1<906, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vctuxs $vD, $vB, $UIMM", IIC_VecFP, + [(set v4i32:$vD, + (int_ppc_altivec_vctuxs v4f32:$vB, imm:$UIMM))]>; + +// Defines with the UIM field set to 0 for floating-point +// to integer (fp_to_sint/fp_to_uint) conversions and integer +// to floating-point (sint_to_fp/uint_to_fp) conversions. +let isCodeGenOnly = 1, VA = 0 in { +def VCFSX_0 : VXForm_1<842, (outs vrrc:$vD), (ins vrrc:$vB), + "vcfsx $vD, $vB, 0", IIC_VecFP, + [(set v4f32:$vD, + (int_ppc_altivec_vcfsx v4i32:$vB, 0))]>; +def VCTUXS_0 : VXForm_1<906, (outs vrrc:$vD), (ins vrrc:$vB), + "vctuxs $vD, $vB, 0", IIC_VecFP, + [(set v4i32:$vD, + (int_ppc_altivec_vctuxs v4f32:$vB, 0))]>; +def VCFUX_0 : VXForm_1<778, (outs vrrc:$vD), (ins vrrc:$vB), + "vcfux $vD, $vB, 0", IIC_VecFP, + [(set v4f32:$vD, + (int_ppc_altivec_vcfux v4i32:$vB, 0))]>; +def VCTSXS_0 : VXForm_1<970, (outs vrrc:$vD), (ins vrrc:$vB), + "vctsxs $vD, $vB, 0", IIC_VecFP, + [(set v4i32:$vD, + (int_ppc_altivec_vctsxs v4f32:$vB, 0))]>; +} +def VEXPTEFP : VX2_Int_SP<394, "vexptefp", int_ppc_altivec_vexptefp>; +def VLOGEFP : VX2_Int_SP<458, "vlogefp", int_ppc_altivec_vlogefp>; + +let isCommutable = 1 in { +def VAVGSB : VX1_Int_Ty<1282, "vavgsb", int_ppc_altivec_vavgsb, v16i8>; +def VAVGSH : VX1_Int_Ty<1346, "vavgsh", int_ppc_altivec_vavgsh, v8i16>; +def VAVGSW : VX1_Int_Ty<1410, "vavgsw", int_ppc_altivec_vavgsw, v4i32>; +def VAVGUB : VX1_Int_Ty<1026, "vavgub", int_ppc_altivec_vavgub, v16i8>; +def VAVGUH : VX1_Int_Ty<1090, "vavguh", int_ppc_altivec_vavguh, v8i16>; +def VAVGUW : VX1_Int_Ty<1154, "vavguw", int_ppc_altivec_vavguw, v4i32>; + +def VMAXFP : VX1_Int_Ty<1034, "vmaxfp", int_ppc_altivec_vmaxfp, v4f32>; +def VMAXSB : VX1_Int_Ty< 258, "vmaxsb", int_ppc_altivec_vmaxsb, v16i8>; +def VMAXSH : VX1_Int_Ty< 322, "vmaxsh", int_ppc_altivec_vmaxsh, v8i16>; +def VMAXSW : VX1_Int_Ty< 386, "vmaxsw", int_ppc_altivec_vmaxsw, v4i32>; +def VMAXUB : VX1_Int_Ty< 2, "vmaxub", int_ppc_altivec_vmaxub, v16i8>; +def VMAXUH : VX1_Int_Ty< 66, "vmaxuh", int_ppc_altivec_vmaxuh, v8i16>; +def VMAXUW : VX1_Int_Ty< 130, "vmaxuw", int_ppc_altivec_vmaxuw, v4i32>; +def VMINFP : VX1_Int_Ty<1098, "vminfp", int_ppc_altivec_vminfp, v4f32>; +def VMINSB : VX1_Int_Ty< 770, "vminsb", int_ppc_altivec_vminsb, v16i8>; +def VMINSH : VX1_Int_Ty< 834, "vminsh", int_ppc_altivec_vminsh, v8i16>; +def VMINSW : VX1_Int_Ty< 898, "vminsw", int_ppc_altivec_vminsw, v4i32>; +def VMINUB : VX1_Int_Ty< 514, "vminub", int_ppc_altivec_vminub, v16i8>; +def VMINUH : VX1_Int_Ty< 578, "vminuh", int_ppc_altivec_vminuh, v8i16>; +def VMINUW : VX1_Int_Ty< 642, "vminuw", int_ppc_altivec_vminuw, v4i32>; +} // isCommutable + +def VMRGHB : VXForm_1< 12, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrghb $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrghb_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGHH : VXForm_1< 76, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrghh $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrghh_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGHW : VXForm_1<140, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrghw $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrghw_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLB : VXForm_1<268, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrglb $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrglb_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLH : VXForm_1<332, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrglh $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrglh_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLW : VXForm_1<396, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vmrglw $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, (vmrglw_shuffle v16i8:$vA, v16i8:$vB))]>; + +def VMSUMMBM : VA1a_Int_Ty3<37, "vmsummbm", int_ppc_altivec_vmsummbm, + v4i32, v16i8, v4i32>; +def VMSUMSHM : VA1a_Int_Ty3<40, "vmsumshm", int_ppc_altivec_vmsumshm, + v4i32, v8i16, v4i32>; +def VMSUMSHS : VA1a_Int_Ty3<41, "vmsumshs", int_ppc_altivec_vmsumshs, + v4i32, v8i16, v4i32>; +def VMSUMUBM : VA1a_Int_Ty3<36, "vmsumubm", int_ppc_altivec_vmsumubm, + v4i32, v16i8, v4i32>; +def VMSUMUHM : VA1a_Int_Ty3<38, "vmsumuhm", int_ppc_altivec_vmsumuhm, + v4i32, v8i16, v4i32>; +def VMSUMUHS : VA1a_Int_Ty3<39, "vmsumuhs", int_ppc_altivec_vmsumuhs, + v4i32, v8i16, v4i32>; + +let isCommutable = 1 in { +def VMULESB : VX1_Int_Ty2<776, "vmulesb", int_ppc_altivec_vmulesb, + v8i16, v16i8>; +def VMULESH : VX1_Int_Ty2<840, "vmulesh", int_ppc_altivec_vmulesh, + v4i32, v8i16>; +def VMULEUB : VX1_Int_Ty2<520, "vmuleub", int_ppc_altivec_vmuleub, + v8i16, v16i8>; +def VMULEUH : VX1_Int_Ty2<584, "vmuleuh", int_ppc_altivec_vmuleuh, + v4i32, v8i16>; +def VMULOSB : VX1_Int_Ty2<264, "vmulosb", int_ppc_altivec_vmulosb, + v8i16, v16i8>; +def VMULOSH : VX1_Int_Ty2<328, "vmulosh", int_ppc_altivec_vmulosh, + v4i32, v8i16>; +def VMULOUB : VX1_Int_Ty2< 8, "vmuloub", int_ppc_altivec_vmuloub, + v8i16, v16i8>; +def VMULOUH : VX1_Int_Ty2< 72, "vmulouh", int_ppc_altivec_vmulouh, + v4i32, v8i16>; +} // isCommutable -def VREFP : VX2_Int<266, "vrefp $vD, $vB", int_ppc_altivec_vrefp>; -def VRFIM : VX2_Int<714, "vrfim $vD, $vB", int_ppc_altivec_vrfim>; -def VRFIN : VX2_Int<522, "vrfin $vD, $vB", int_ppc_altivec_vrfin>; -def VRFIP : VX2_Int<650, "vrfip $vD, $vB", int_ppc_altivec_vrfip>; -def VRFIZ : VX2_Int<586, "vrfiz $vD, $vB", int_ppc_altivec_vrfiz>; -def VRSQRTEFP : VX2_Int<330, "vrsqrtefp $vD, $vB", int_ppc_altivec_vrsqrtefp>; - -def VSUBCUW : VX1_Int<74, "vsubcuw $vD, $vA, $vB", int_ppc_altivec_vsubcuw>; - -def VSUBFP : VXForm_1<74, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vsubfp $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (fsub VRRC:$vA, VRRC:$vB))]>; -def VSUBUBM : VXForm_1<1024, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vsububm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VSUBUHM : VXForm_1<1088, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vsubuhm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v8i16 VRRC:$vA), VRRC:$vB))]>; -def VSUBUWM : VXForm_1<1152, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vsubuwm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v4i32 VRRC:$vA), VRRC:$vB))]>; +def VREFP : VX2_Int_SP<266, "vrefp", int_ppc_altivec_vrefp>; +def VRFIM : VX2_Int_SP<714, "vrfim", int_ppc_altivec_vrfim>; +def VRFIN : VX2_Int_SP<522, "vrfin", int_ppc_altivec_vrfin>; +def VRFIP : VX2_Int_SP<650, "vrfip", int_ppc_altivec_vrfip>; +def VRFIZ : VX2_Int_SP<586, "vrfiz", int_ppc_altivec_vrfiz>; +def VRSQRTEFP : VX2_Int_SP<330, "vrsqrtefp", int_ppc_altivec_vrsqrtefp>; + +def VSUBCUW : VX1_Int_Ty<1408, "vsubcuw", int_ppc_altivec_vsubcuw, v4i32>; + +def VSUBFP : VXForm_1<74, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vsubfp $vD, $vA, $vB", IIC_VecGeneral, + [(set v4f32:$vD, (fsub v4f32:$vA, v4f32:$vB))]>; +def VSUBUBM : VXForm_1<1024, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vsububm $vD, $vA, $vB", IIC_VecGeneral, + [(set v16i8:$vD, (sub v16i8:$vA, v16i8:$vB))]>; +def VSUBUHM : VXForm_1<1088, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vsubuhm $vD, $vA, $vB", IIC_VecGeneral, + [(set v8i16:$vD, (sub v8i16:$vA, v8i16:$vB))]>; +def VSUBUWM : VXForm_1<1152, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vsubuwm $vD, $vA, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (sub v4i32:$vA, v4i32:$vB))]>; -def VSUBSBS : VX1_Int<1792, "vsubsbs $vD, $vA, $vB", int_ppc_altivec_vsubsbs>; -def VSUBSHS : VX1_Int<1856, "vsubshs $vD, $vA, $vB", int_ppc_altivec_vsubshs>; -def VSUBSWS : VX1_Int<1920, "vsubsws $vD, $vA, $vB", int_ppc_altivec_vsubsws>; -def VSUBUBS : VX1_Int<1536, "vsububs $vD, $vA, $vB", int_ppc_altivec_vsububs>; -def VSUBUHS : VX1_Int<1600, "vsubuhs $vD, $vA, $vB", int_ppc_altivec_vsubuhs>; -def VSUBUWS : VX1_Int<1664, "vsubuws $vD, $vA, $vB", int_ppc_altivec_vsubuws>; -def VSUMSWS : VX1_Int<1928, "vsumsws $vD, $vA, $vB", int_ppc_altivec_vsumsws>; -def VSUM2SWS: VX1_Int<1672, "vsum2sws $vD, $vA, $vB", int_ppc_altivec_vsum2sws>; -def VSUM4SBS: VX1_Int<1672, "vsum4sbs $vD, $vA, $vB", int_ppc_altivec_vsum4sbs>; -def VSUM4SHS: VX1_Int<1608, "vsum4shs $vD, $vA, $vB", int_ppc_altivec_vsum4shs>; -def VSUM4UBS: VX1_Int<1544, "vsum4ubs $vD, $vA, $vB", int_ppc_altivec_vsum4ubs>; - -def VNOR : VXForm_1<1284, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vnor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vnot (or (v4i32 VRRC:$vA), VRRC:$vB)))]>; -def VOR : VXForm_1<1156, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (or (v4i32 VRRC:$vA), VRRC:$vB))]>; -def VXOR : VXForm_1<1220, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vxor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (xor (v4i32 VRRC:$vA), VRRC:$vB))]>; - -def VRLB : VX1_Int< 4, "vrlb $vD, $vA, $vB", int_ppc_altivec_vrlb>; -def VRLH : VX1_Int< 68, "vrlh $vD, $vA, $vB", int_ppc_altivec_vrlh>; -def VRLW : VX1_Int< 132, "vrlw $vD, $vA, $vB", int_ppc_altivec_vrlw>; -def VSLO : VX1_Int<1036, "vslo $vD, $vA, $vB", int_ppc_altivec_vslo>; -def VSLB : VX1_Int< 260, "vslb $vD, $vA, $vB", int_ppc_altivec_vslb>; -def VSLH : VX1_Int< 324, "vslh $vD, $vA, $vB", int_ppc_altivec_vslh>; -def VSLW : VX1_Int< 388, "vslw $vD, $vA, $vB", int_ppc_altivec_vslw>; - -def VSPLTB : VXForm_1<524, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vspltb $vD, $vB, $UIMM", VecPerm, - []>; -def VSPLTH : VXForm_1<588, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vsplth $vD, $vB, $UIMM", VecPerm, - []>; -def VSPLTW : VXForm_1<652, (ops VRRC:$vD, u5imm:$UIMM, VRRC:$vB), - "vspltw $vD, $vB, $UIMM", VecPerm, - [(set VRRC:$vD, (vector_shuffle (v4f32 VRRC:$vB), (undef), - VSPLT_shuffle_mask:$UIMM))]>; - -def VSR : VX1_Int< 708, "vsr $vD, $vA, $vB" , int_ppc_altivec_vsr>; -def VSRO : VX1_Int<1100, "vsro $vD, $vA, $vB" , int_ppc_altivec_vsro>; -def VSRAB : VX1_Int< 772, "vsrab $vD, $vA, $vB", int_ppc_altivec_vsrab>; -def VSRAH : VX1_Int< 836, "vsrah $vD, $vA, $vB", int_ppc_altivec_vsrah>; -def VSRAW : VX1_Int< 900, "vsraw $vD, $vA, $vB", int_ppc_altivec_vsraw>; -def VSRB : VX1_Int< 516, "vsrb $vD, $vA, $vB" , int_ppc_altivec_vsrb>; -def VSRH : VX1_Int< 580, "vsrh $vD, $vA, $vB" , int_ppc_altivec_vsrh>; -def VSRW : VX1_Int< 644, "vsrw $vD, $vA, $vB" , int_ppc_altivec_vsrw>; - - -def VSPLTISB : VXForm_3<780, (ops VRRC:$vD, s5imm:$SIMM), - "vspltisb $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v4f32 vecspltisb:$SIMM))]>; -def VSPLTISH : VXForm_3<844, (ops VRRC:$vD, s5imm:$SIMM), - "vspltish $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v4f32 vecspltish:$SIMM))]>; -def VSPLTISW : VXForm_3<908, (ops VRRC:$vD, s5imm:$SIMM), - "vspltisw $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v4f32 vecspltisw:$SIMM))]>; +def VSUBSBS : VX1_Int_Ty<1792, "vsubsbs" , int_ppc_altivec_vsubsbs, v16i8>; +def VSUBSHS : VX1_Int_Ty<1856, "vsubshs" , int_ppc_altivec_vsubshs, v8i16>; +def VSUBSWS : VX1_Int_Ty<1920, "vsubsws" , int_ppc_altivec_vsubsws, v4i32>; +def VSUBUBS : VX1_Int_Ty<1536, "vsububs" , int_ppc_altivec_vsububs, v16i8>; +def VSUBUHS : VX1_Int_Ty<1600, "vsubuhs" , int_ppc_altivec_vsubuhs, v8i16>; +def VSUBUWS : VX1_Int_Ty<1664, "vsubuws" , int_ppc_altivec_vsubuws, v4i32>; + +def VSUMSWS : VX1_Int_Ty<1928, "vsumsws" , int_ppc_altivec_vsumsws, v4i32>; +def VSUM2SWS: VX1_Int_Ty<1672, "vsum2sws", int_ppc_altivec_vsum2sws, v4i32>; + +def VSUM4SBS: VX1_Int_Ty3<1800, "vsum4sbs", int_ppc_altivec_vsum4sbs, + v4i32, v16i8, v4i32>; +def VSUM4SHS: VX1_Int_Ty3<1608, "vsum4shs", int_ppc_altivec_vsum4shs, + v4i32, v8i16, v4i32>; +def VSUM4UBS: VX1_Int_Ty3<1544, "vsum4ubs", int_ppc_altivec_vsum4ubs, + v4i32, v16i8, v4i32>; + +def VNOR : VXForm_1<1284, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vnor $vD, $vA, $vB", IIC_VecFP, + [(set v4i32:$vD, (vnot_ppc (or v4i32:$vA, + v4i32:$vB)))]>; +let isCommutable = 1 in { +def VOR : VXForm_1<1156, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vor $vD, $vA, $vB", IIC_VecFP, + [(set v4i32:$vD, (or v4i32:$vA, v4i32:$vB))]>; +def VXOR : VXForm_1<1220, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vxor $vD, $vA, $vB", IIC_VecFP, + [(set v4i32:$vD, (xor v4i32:$vA, v4i32:$vB))]>; +} // isCommutable + +def VRLB : VX1_Int_Ty< 4, "vrlb", int_ppc_altivec_vrlb, v16i8>; +def VRLH : VX1_Int_Ty< 68, "vrlh", int_ppc_altivec_vrlh, v8i16>; +def VRLW : VX1_Int_Ty< 132, "vrlw", int_ppc_altivec_vrlw, v4i32>; + +def VSL : VX1_Int_Ty< 452, "vsl" , int_ppc_altivec_vsl, v4i32 >; +def VSLO : VX1_Int_Ty<1036, "vslo", int_ppc_altivec_vslo, v4i32>; + +def VSLB : VX1_Int_Ty< 260, "vslb", int_ppc_altivec_vslb, v16i8>; +def VSLH : VX1_Int_Ty< 324, "vslh", int_ppc_altivec_vslh, v8i16>; +def VSLW : VX1_Int_Ty< 388, "vslw", int_ppc_altivec_vslw, v4i32>; + +def VSPLTB : VXForm_1<524, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vspltb $vD, $vB, $UIMM", IIC_VecPerm, + [(set v16i8:$vD, + (vspltb_shuffle:$UIMM v16i8:$vB, (undef)))]>; +def VSPLTH : VXForm_1<588, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vsplth $vD, $vB, $UIMM", IIC_VecPerm, + [(set v16i8:$vD, + (vsplth_shuffle:$UIMM v16i8:$vB, (undef)))]>; +def VSPLTW : VXForm_1<652, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), + "vspltw $vD, $vB, $UIMM", IIC_VecPerm, + [(set v16i8:$vD, + (vspltw_shuffle:$UIMM v16i8:$vB, (undef)))]>; + +def VSR : VX1_Int_Ty< 708, "vsr" , int_ppc_altivec_vsr, v4i32>; +def VSRO : VX1_Int_Ty<1100, "vsro" , int_ppc_altivec_vsro, v4i32>; + +def VSRAB : VX1_Int_Ty< 772, "vsrab", int_ppc_altivec_vsrab, v16i8>; +def VSRAH : VX1_Int_Ty< 836, "vsrah", int_ppc_altivec_vsrah, v8i16>; +def VSRAW : VX1_Int_Ty< 900, "vsraw", int_ppc_altivec_vsraw, v4i32>; +def VSRB : VX1_Int_Ty< 516, "vsrb" , int_ppc_altivec_vsrb , v16i8>; +def VSRH : VX1_Int_Ty< 580, "vsrh" , int_ppc_altivec_vsrh , v8i16>; +def VSRW : VX1_Int_Ty< 644, "vsrw" , int_ppc_altivec_vsrw , v4i32>; + + +def VSPLTISB : VXForm_3<780, (outs vrrc:$vD), (ins s5imm:$SIMM), + "vspltisb $vD, $SIMM", IIC_VecPerm, + [(set v16i8:$vD, (v16i8 vecspltisb:$SIMM))]>; +def VSPLTISH : VXForm_3<844, (outs vrrc:$vD), (ins s5imm:$SIMM), + "vspltish $vD, $SIMM", IIC_VecPerm, + [(set v8i16:$vD, (v8i16 vecspltish:$SIMM))]>; +def VSPLTISW : VXForm_3<908, (outs vrrc:$vD), (ins s5imm:$SIMM), + "vspltisw $vD, $SIMM", IIC_VecPerm, + [(set v4i32:$vD, (v4i32 vecspltisw:$SIMM))]>; // Vector Pack. -def VPKPX : VXForm_1<782, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkpx $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkpx VRRC:$vA, VRRC:$vB))]>; -def VPKSHSS : VXForm_1<398, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkshss $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkshss VRRC:$vA, VRRC:$vB))]>; -def VPKSHUS : VXForm_1<270, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkshus $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkshus VRRC:$vA, VRRC:$vB))]>; -def VPKSWSS : VXForm_1<462, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkswss $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkswss VRRC:$vA, VRRC:$vB))]>; -def VPKSWUS : VXForm_1<334, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkswus $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkswus VRRC:$vA, VRRC:$vB))]>; -def VPKUHUM : VXForm_1<14, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkuhum $vD, $vA, $vB", VecFP, - [/*TODO*/]>; -def VPKUHUS : VXForm_1<142, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkuhus $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkuhus VRRC:$vA, VRRC:$vB))]>; -def VPKUWUM : VXForm_1<78, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkuwum $vD, $vA, $vB", VecFP, - [/*TODO*/]>; -def VPKUWUS : VXForm_1<206, (ops VRRC:$vD, VRRC:$vA, VRRC:$vB), - "vpkuwus $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vpkuwus VRRC:$vA, VRRC:$vB))]>; +def VPKPX : VX1_Int_Ty2<782, "vpkpx", int_ppc_altivec_vpkpx, + v8i16, v4i32>; +def VPKSHSS : VX1_Int_Ty2<398, "vpkshss", int_ppc_altivec_vpkshss, + v16i8, v8i16>; +def VPKSHUS : VX1_Int_Ty2<270, "vpkshus", int_ppc_altivec_vpkshus, + v16i8, v8i16>; +def VPKSWSS : VX1_Int_Ty2<462, "vpkswss", int_ppc_altivec_vpkswss, + v16i8, v4i32>; +def VPKSWUS : VX1_Int_Ty2<334, "vpkswus", int_ppc_altivec_vpkswus, + v8i16, v4i32>; +def VPKUHUM : VXForm_1<14, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vpkuhum $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, + (vpkuhum_shuffle v16i8:$vA, v16i8:$vB))]>; +def VPKUHUS : VX1_Int_Ty2<142, "vpkuhus", int_ppc_altivec_vpkuhus, + v16i8, v8i16>; +def VPKUWUM : VXForm_1<78, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vpkuwum $vD, $vA, $vB", IIC_VecFP, + [(set v16i8:$vD, + (vpkuwum_shuffle v16i8:$vA, v16i8:$vB))]>; +def VPKUWUS : VX1_Int_Ty2<206, "vpkuwus", int_ppc_altivec_vpkuwus, + v8i16, v4i32>; // Vector Unpack. -def VUPKHPX : VXForm_2<846, (ops VRRC:$vD, VRRC:$vB), - "vupkhpx $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupkhpx VRRC:$vB))]>; -def VUPKHSB : VXForm_2<526, (ops VRRC:$vD, VRRC:$vB), - "vupkhsb $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupkhsb VRRC:$vB))]>; -def VUPKHSH : VXForm_2<590, (ops VRRC:$vD, VRRC:$vB), - "vupkhsh $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupkhsh VRRC:$vB))]>; -def VUPKLPX : VXForm_2<974, (ops VRRC:$vD, VRRC:$vB), - "vupklpx $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupklpx VRRC:$vB))]>; -def VUPKLSB : VXForm_2<654, (ops VRRC:$vD, VRRC:$vB), - "vupklsb $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupklsb VRRC:$vB))]>; -def VUPKLSH : VXForm_2<718, (ops VRRC:$vD, VRRC:$vB), - "vupklsh $vD, $vB", VecFP, - [(set VRRC:$vD, (int_ppc_altivec_vupklsh VRRC:$vB))]>; +def VUPKHPX : VX2_Int_Ty2<846, "vupkhpx", int_ppc_altivec_vupkhpx, + v4i32, v8i16>; +def VUPKHSB : VX2_Int_Ty2<526, "vupkhsb", int_ppc_altivec_vupkhsb, + v8i16, v16i8>; +def VUPKHSH : VX2_Int_Ty2<590, "vupkhsh", int_ppc_altivec_vupkhsh, + v4i32, v8i16>; +def VUPKLPX : VX2_Int_Ty2<974, "vupklpx", int_ppc_altivec_vupklpx, + v4i32, v8i16>; +def VUPKLSB : VX2_Int_Ty2<654, "vupklsb", int_ppc_altivec_vupklsb, + v8i16, v16i8>; +def VUPKLSH : VX2_Int_Ty2<718, "vupklsh", int_ppc_altivec_vupklsh, + v4i32, v8i16>; // Altivec Comparisons. class VCMP xo, string asmstr, ValueType Ty> - : VXRForm_1; + : VXRForm_1; class VCMPo xo, string asmstr, ValueType Ty> - : VXRForm_1,isVDOT; + : VXRForm_1 { + let Defs = [CR6]; + let RC = 1; +} // f32 element comparisons.0 def VCMPBFP : VCMP <966, "vcmpbfp $vD, $vA, $vB" , v4f32>; @@ -434,107 +760,319 @@ def VCMPGTSW : VCMP <902, "vcmpgtsw $vD, $vA, $vB" , v4i32>; def VCMPGTSWo : VCMPo<902, "vcmpgtsw. $vD, $vA, $vB", v4i32>; def VCMPGTUW : VCMP <646, "vcmpgtuw $vD, $vA, $vB" , v4i32>; def VCMPGTUWo : VCMPo<646, "vcmpgtuw. $vD, $vA, $vB", v4i32>; - -def V_SET0 : VXForm_setzero<1220, (ops VRRC:$vD), - "vxor $vD, $vD, $vD", VecFP, - [(set VRRC:$vD, (v4f32 immAllZerosV))]>; + +let isCodeGenOnly = 1 in { +def V_SET0B : VXForm_setzero<1220, (outs vrrc:$vD), (ins), + "vxor $vD, $vD, $vD", IIC_VecFP, + [(set v16i8:$vD, (v16i8 immAllZerosV))]>; +def V_SET0H : VXForm_setzero<1220, (outs vrrc:$vD), (ins), + "vxor $vD, $vD, $vD", IIC_VecFP, + [(set v8i16:$vD, (v8i16 immAllZerosV))]>; +def V_SET0 : VXForm_setzero<1220, (outs vrrc:$vD), (ins), + "vxor $vD, $vD, $vD", IIC_VecFP, + [(set v4i32:$vD, (v4i32 immAllZerosV))]>; + +let IMM=-1 in { +def V_SETALLONESB : VXForm_3<908, (outs vrrc:$vD), (ins), + "vspltisw $vD, -1", IIC_VecFP, + [(set v16i8:$vD, (v16i8 immAllOnesV))]>; +def V_SETALLONESH : VXForm_3<908, (outs vrrc:$vD), (ins), + "vspltisw $vD, -1", IIC_VecFP, + [(set v8i16:$vD, (v8i16 immAllOnesV))]>; +def V_SETALLONES : VXForm_3<908, (outs vrrc:$vD), (ins), + "vspltisw $vD, -1", IIC_VecFP, + [(set v4i32:$vD, (v4i32 immAllOnesV))]>; } +} +} // VALU Operations. //===----------------------------------------------------------------------===// // Additional Altivec Patterns // -// Undef/Zero. -def : Pat<(v16i8 (undef)), (v16i8 (IMPLICIT_DEF_VRRC))>; -def : Pat<(v8i16 (undef)), (v8i16 (IMPLICIT_DEF_VRRC))>; -def : Pat<(v4i32 (undef)), (v4i32 (IMPLICIT_DEF_VRRC))>; -def : Pat<(v16i8 immAllZerosV), (v16i8 (V_SET0))>; -def : Pat<(v8i16 immAllZerosV), (v8i16 (V_SET0))>; -def : Pat<(v4i32 immAllZerosV), (v4i32 (V_SET0))>; - // Loads. -def : Pat<(v16i8 (load xoaddr:$src)), (v16i8 (LVX xoaddr:$src))>; -def : Pat<(v8i16 (load xoaddr:$src)), (v8i16 (LVX xoaddr:$src))>; -def : Pat<(v4i32 (load xoaddr:$src)), (v4i32 (LVX xoaddr:$src))>; -def : Pat<(v4f32 (load xoaddr:$src)), (v4f32 (LVX xoaddr:$src))>; +def : Pat<(v4i32 (load xoaddr:$src)), (LVX xoaddr:$src)>; // Stores. -def : Pat<(store (v16i8 VRRC:$rS), xoaddr:$dst), - (STVX (v16i8 VRRC:$rS), xoaddr:$dst)>; -def : Pat<(store (v8i16 VRRC:$rS), xoaddr:$dst), - (STVX (v8i16 VRRC:$rS), xoaddr:$dst)>; -def : Pat<(store (v4i32 VRRC:$rS), xoaddr:$dst), - (STVX (v4i32 VRRC:$rS), xoaddr:$dst)>; -def : Pat<(store (v4f32 VRRC:$rS), xoaddr:$dst), - (STVX (v4f32 VRRC:$rS), xoaddr:$dst)>; +def : Pat<(store v4i32:$rS, xoaddr:$dst), + (STVX $rS, xoaddr:$dst)>; // Bit conversions. def : Pat<(v16i8 (bitconvert (v8i16 VRRC:$src))), (v16i8 VRRC:$src)>; def : Pat<(v16i8 (bitconvert (v4i32 VRRC:$src))), (v16i8 VRRC:$src)>; def : Pat<(v16i8 (bitconvert (v4f32 VRRC:$src))), (v16i8 VRRC:$src)>; +def : Pat<(v16i8 (bitconvert (v2i64 VRRC:$src))), (v16i8 VRRC:$src)>; def : Pat<(v8i16 (bitconvert (v16i8 VRRC:$src))), (v8i16 VRRC:$src)>; def : Pat<(v8i16 (bitconvert (v4i32 VRRC:$src))), (v8i16 VRRC:$src)>; def : Pat<(v8i16 (bitconvert (v4f32 VRRC:$src))), (v8i16 VRRC:$src)>; +def : Pat<(v8i16 (bitconvert (v2i64 VRRC:$src))), (v8i16 VRRC:$src)>; def : Pat<(v4i32 (bitconvert (v16i8 VRRC:$src))), (v4i32 VRRC:$src)>; def : Pat<(v4i32 (bitconvert (v8i16 VRRC:$src))), (v4i32 VRRC:$src)>; def : Pat<(v4i32 (bitconvert (v4f32 VRRC:$src))), (v4i32 VRRC:$src)>; +def : Pat<(v4i32 (bitconvert (v2i64 VRRC:$src))), (v4i32 VRRC:$src)>; def : Pat<(v4f32 (bitconvert (v16i8 VRRC:$src))), (v4f32 VRRC:$src)>; def : Pat<(v4f32 (bitconvert (v8i16 VRRC:$src))), (v4f32 VRRC:$src)>; def : Pat<(v4f32 (bitconvert (v4i32 VRRC:$src))), (v4f32 VRRC:$src)>; +def : Pat<(v4f32 (bitconvert (v2i64 VRRC:$src))), (v4f32 VRRC:$src)>; -// Immediate vector formation with vsplti*. -def : Pat<(v16i8 vecspltisb:$invec), (v16i8 (VSPLTISB vecspltisb:$invec))>; -def : Pat<(v16i8 vecspltish:$invec), (v16i8 (VSPLTISH vecspltish:$invec))>; -def : Pat<(v16i8 vecspltisw:$invec), (v16i8 (VSPLTISW vecspltisw:$invec))>; +def : Pat<(v2i64 (bitconvert (v16i8 VRRC:$src))), (v2i64 VRRC:$src)>; +def : Pat<(v2i64 (bitconvert (v8i16 VRRC:$src))), (v2i64 VRRC:$src)>; +def : Pat<(v2i64 (bitconvert (v4i32 VRRC:$src))), (v2i64 VRRC:$src)>; +def : Pat<(v2i64 (bitconvert (v4f32 VRRC:$src))), (v2i64 VRRC:$src)>; -def : Pat<(v8i16 vecspltisb:$invec), (v8i16 (VSPLTISB vecspltisb:$invec))>; -def : Pat<(v8i16 vecspltish:$invec), (v8i16 (VSPLTISH vecspltish:$invec))>; -def : Pat<(v8i16 vecspltisw:$invec), (v8i16 (VSPLTISW vecspltisw:$invec))>; +// Shuffles. -def : Pat<(v4i32 vecspltisb:$invec), (v4i32 (VSPLTISB vecspltisb:$invec))>; -def : Pat<(v4i32 vecspltish:$invec), (v4i32 (VSPLTISH vecspltish:$invec))>; -def : Pat<(v4i32 vecspltisw:$invec), (v4i32 (VSPLTISW vecspltisw:$invec))>; +// Match vsldoi(x,x), vpkuwum(x,x), vpkuhum(x,x) +def:Pat<(vsldoi_unary_shuffle:$in v16i8:$vA, undef), + (VSLDOI $vA, $vA, (VSLDOI_unary_get_imm $in))>; +def:Pat<(vpkuwum_unary_shuffle v16i8:$vA, undef), + (VPKUWUM $vA, $vA)>; +def:Pat<(vpkuhum_unary_shuffle v16i8:$vA, undef), + (VPKUHUM $vA, $vA)>; + +// Match vsldoi(y,x), vpkuwum(y,x), vpkuhum(y,x), i.e., swapped operands. +// These fragments are matched for little-endian, where the inputs must +// be swapped for correct semantics. +def:Pat<(vsldoi_swapped_shuffle:$in v16i8:$vA, v16i8:$vB), + (VSLDOI $vB, $vA, (VSLDOI_swapped_get_imm $in))>; +def:Pat<(vpkuwum_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VPKUWUM $vB, $vA)>; +def:Pat<(vpkuhum_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VPKUHUM $vB, $vA)>; + +// Match vmrg*(x,x) +def:Pat<(vmrglb_unary_shuffle v16i8:$vA, undef), + (VMRGLB $vA, $vA)>; +def:Pat<(vmrglh_unary_shuffle v16i8:$vA, undef), + (VMRGLH $vA, $vA)>; +def:Pat<(vmrglw_unary_shuffle v16i8:$vA, undef), + (VMRGLW $vA, $vA)>; +def:Pat<(vmrghb_unary_shuffle v16i8:$vA, undef), + (VMRGHB $vA, $vA)>; +def:Pat<(vmrghh_unary_shuffle v16i8:$vA, undef), + (VMRGHH $vA, $vA)>; +def:Pat<(vmrghw_unary_shuffle v16i8:$vA, undef), + (VMRGHW $vA, $vA)>; + +// Match vmrg*(y,x), i.e., swapped operands. These fragments +// are matched for little-endian, where the inputs must be +// swapped for correct semantics. +def:Pat<(vmrglb_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGLB $vB, $vA)>; +def:Pat<(vmrglh_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGLH $vB, $vA)>; +def:Pat<(vmrglw_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGLW $vB, $vA)>; +def:Pat<(vmrghb_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGHB $vB, $vA)>; +def:Pat<(vmrghh_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGHH $vB, $vA)>; +def:Pat<(vmrghw_swapped_shuffle v16i8:$vA, v16i8:$vB), + (VMRGHW $vB, $vA)>; // Logical Operations -def : Pat<(v16i8 (and VRRC:$A, VRRC:$B)), (v16i8 (VAND VRRC:$A, VRRC:$B))>; -def : Pat<(v8i16 (and VRRC:$A, VRRC:$B)), (v8i16 (VAND VRRC:$A, VRRC:$B))>; -def : Pat<(v16i8 (or VRRC:$A, VRRC:$B)), (v16i8 (VOR VRRC:$A, VRRC:$B))>; -def : Pat<(v8i16 (or VRRC:$A, VRRC:$B)), (v8i16 (VOR VRRC:$A, VRRC:$B))>; -def : Pat<(v16i8 (xor VRRC:$A, VRRC:$B)), (v16i8 (VXOR VRRC:$A, VRRC:$B))>; -def : Pat<(v8i16 (xor VRRC:$A, VRRC:$B)), (v8i16 (VXOR VRRC:$A, VRRC:$B))>; -def : Pat<(v16i8 (vnot (or VRRC:$A, VRRC:$B))),(v16i8 (VNOR VRRC:$A, VRRC:$B))>; -def : Pat<(v8i16 (vnot (or VRRC:$A, VRRC:$B))),(v8i16 (VNOR VRRC:$A, VRRC:$B))>; -def : Pat<(v16i8 (and VRRC:$A, (vnot VRRC:$B))), - (v16i8 (VANDC VRRC:$A, VRRC:$B))>; -def : Pat<(v8i16 (and VRRC:$A, (vnot VRRC:$B))), - (v8i16 (VANDC VRRC:$A, VRRC:$B))>; - -def : Pat<(fmul VRRC:$vA, VRRC:$vB), - (VMADDFP VRRC:$vA, VRRC:$vB, (V_SET0))>; +def : Pat<(vnot_ppc v4i32:$vA), (VNOR $vA, $vA)>; + +def : Pat<(vnot_ppc (or v4i32:$A, v4i32:$B)), + (VNOR $A, $B)>; +def : Pat<(and v4i32:$A, (vnot_ppc v4i32:$B)), + (VANDC $A, $B)>; + +def : Pat<(fmul v4f32:$vA, v4f32:$vB), + (VMADDFP $vA, $vB, + (v4i32 (VSLW (V_SETALLONES), (V_SETALLONES))))>; // Fused multiply add and multiply sub for packed float. These are represented // separately from the real instructions above, for operations that must have // the additional precision, such as Newton-Rhapson (used by divide, sqrt) -def : Pat<(PPCvmaddfp VRRC:$A, VRRC:$B, VRRC:$C), - (VMADDFP VRRC:$A, VRRC:$B, VRRC:$C)>; -def : Pat<(PPCvnmsubfp VRRC:$A, VRRC:$B, VRRC:$C), - (VNMSUBFP VRRC:$A, VRRC:$B, VRRC:$C)>; - -def : Pat<(int_ppc_altivec_vmaddfp VRRC:$A, VRRC:$B, VRRC:$C), - (VMADDFP VRRC:$A, VRRC:$B, VRRC:$C)>; -def : Pat<(int_ppc_altivec_vnmsubfp VRRC:$A, VRRC:$B, VRRC:$C), - (VNMSUBFP VRRC:$A, VRRC:$B, VRRC:$C)>; -def : Pat<(vector_shuffle (v4i32 VRRC:$vB), (undef), VSPLT_shuffle_mask:$UIMM), - (v4i32 (VSPLTW VSPLT_shuffle_mask:$UIMM, VRRC:$vB))>; - -def : Pat<(PPCvperm (v4i32 VRRC:$vA), VRRC:$vB, VRRC:$vC), - (v4i32 (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC))>; -def : Pat<(PPCvperm (v4f32 VRRC:$vA), VRRC:$vB, VRRC:$vC), - (v4f32 (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC))>; -def : Pat<(PPCvperm (v8i16 VRRC:$vA), VRRC:$vB, VRRC:$vC), - (v8i16 (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC))>; -def : Pat<(PPCvperm (v16i8 VRRC:$vA), VRRC:$vB, VRRC:$vC), - (v16i8 (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC))>; +def : Pat<(PPCvmaddfp v4f32:$A, v4f32:$B, v4f32:$C), + (VMADDFP $A, $B, $C)>; +def : Pat<(PPCvnmsubfp v4f32:$A, v4f32:$B, v4f32:$C), + (VNMSUBFP $A, $B, $C)>; + +def : Pat<(int_ppc_altivec_vmaddfp v4f32:$A, v4f32:$B, v4f32:$C), + (VMADDFP $A, $B, $C)>; +def : Pat<(int_ppc_altivec_vnmsubfp v4f32:$A, v4f32:$B, v4f32:$C), + (VNMSUBFP $A, $B, $C)>; + +def : Pat<(PPCvperm v16i8:$vA, v16i8:$vB, v16i8:$vC), + (VPERM $vA, $vB, $vC)>; + +def : Pat<(PPCfre v4f32:$A), (VREFP $A)>; +def : Pat<(PPCfrsqrte v4f32:$A), (VRSQRTEFP $A)>; + +// Vector shifts +def : Pat<(v16i8 (shl v16i8:$vA, v16i8:$vB)), + (v16i8 (VSLB $vA, $vB))>; +def : Pat<(v8i16 (shl v8i16:$vA, v8i16:$vB)), + (v8i16 (VSLH $vA, $vB))>; +def : Pat<(v4i32 (shl v4i32:$vA, v4i32:$vB)), + (v4i32 (VSLW $vA, $vB))>; + +def : Pat<(v16i8 (srl v16i8:$vA, v16i8:$vB)), + (v16i8 (VSRB $vA, $vB))>; +def : Pat<(v8i16 (srl v8i16:$vA, v8i16:$vB)), + (v8i16 (VSRH $vA, $vB))>; +def : Pat<(v4i32 (srl v4i32:$vA, v4i32:$vB)), + (v4i32 (VSRW $vA, $vB))>; + +def : Pat<(v16i8 (sra v16i8:$vA, v16i8:$vB)), + (v16i8 (VSRAB $vA, $vB))>; +def : Pat<(v8i16 (sra v8i16:$vA, v8i16:$vB)), + (v8i16 (VSRAH $vA, $vB))>; +def : Pat<(v4i32 (sra v4i32:$vA, v4i32:$vB)), + (v4i32 (VSRAW $vA, $vB))>; + +// Float to integer and integer to float conversions +def : Pat<(v4i32 (fp_to_sint v4f32:$vA)), + (VCTSXS_0 $vA)>; +def : Pat<(v4i32 (fp_to_uint v4f32:$vA)), + (VCTUXS_0 $vA)>; +def : Pat<(v4f32 (sint_to_fp v4i32:$vA)), + (VCFSX_0 $vA)>; +def : Pat<(v4f32 (uint_to_fp v4i32:$vA)), + (VCFUX_0 $vA)>; + +// Floating-point rounding +def : Pat<(v4f32 (ffloor v4f32:$vA)), + (VRFIM $vA)>; +def : Pat<(v4f32 (fceil v4f32:$vA)), + (VRFIP $vA)>; +def : Pat<(v4f32 (ftrunc v4f32:$vA)), + (VRFIZ $vA)>; +def : Pat<(v4f32 (fnearbyint v4f32:$vA)), + (VRFIN $vA)>; + +} // end HasAltivec + +def HasP8Altivec : Predicate<"PPCSubTarget->hasP8Altivec()">; +def HasP8Crypto : Predicate<"PPCSubTarget->hasP8Crypto()">; +let Predicates = [HasP8Altivec] in { + +let isCommutable = 1 in { +def VMULESW : VX1_Int_Ty2<904, "vmulesw", int_ppc_altivec_vmulesw, + v2i64, v4i32>; +def VMULEUW : VX1_Int_Ty2<648, "vmuleuw", int_ppc_altivec_vmuleuw, + v2i64, v4i32>; +def VMULOSW : VX1_Int_Ty2<392, "vmulosw", int_ppc_altivec_vmulosw, + v2i64, v4i32>; +def VMULOUW : VX1_Int_Ty2<136, "vmulouw", int_ppc_altivec_vmulouw, + v2i64, v4i32>; +def VMULUWM : VX1_Int_Ty<137, "vmuluwm", int_ppc_altivec_vmuluwm, + v4i32>; +def VMAXSD : VX1_Int_Ty<450, "vmaxsd", int_ppc_altivec_vmaxsd, v2i64>; +def VMAXUD : VX1_Int_Ty<194, "vmaxud", int_ppc_altivec_vmaxud, v2i64>; +def VMINSD : VX1_Int_Ty<962, "vminsd", int_ppc_altivec_vminsd, v2i64>; +def VMIDUD : VX1_Int_Ty<706, "vminud", int_ppc_altivec_vminud, v2i64>; +} // isCommutable + +def VRLD : VX1_Int_Ty<196, "vrld", int_ppc_altivec_vrld, v2i64>; +def VSLD : VX1_Int_Ty<1476, "vsld", int_ppc_altivec_vsld, v2i64>; +def VSRD : VX1_Int_Ty<1732, "vsrd", int_ppc_altivec_vsrd, v2i64>; +def VSRAD : VX1_Int_Ty<964, "vsrad", int_ppc_altivec_vsrad, v2i64>; + + +// Vector Integer Arithmetic Instructions +let isCommutable = 1 in { +def VADDUDM : VXForm_1<192, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vaddudm $vD, $vA, $vB", IIC_VecGeneral, + [(set v2i64:$vD, (add v2i64:$vA, v2i64:$vB))]>; +} // isCommutable + +def VSUBUDM : VXForm_1<1216, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vsubudm $vD, $vA, $vB", IIC_VecGeneral, + [(set v2i64:$vD, (sub v2i64:$vA, v2i64:$vB))]>; + +// Count Leading Zeros +def VCLZB : VXForm_2<1794, (outs vrrc:$vD), (ins vrrc:$vB), + "vclzb $vD, $vB", IIC_VecGeneral, + [(set v16i8:$vD, (ctlz v16i8:$vB))]>; +def VCLZH : VXForm_2<1858, (outs vrrc:$vD), (ins vrrc:$vB), + "vclzh $vD, $vB", IIC_VecGeneral, + [(set v8i16:$vD, (ctlz v8i16:$vB))]>; +def VCLZW : VXForm_2<1922, (outs vrrc:$vD), (ins vrrc:$vB), + "vclzw $vD, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (ctlz v4i32:$vB))]>; +def VCLZD : VXForm_2<1986, (outs vrrc:$vD), (ins vrrc:$vB), + "vclzd $vD, $vB", IIC_VecGeneral, + [(set v2i64:$vD, (ctlz v2i64:$vB))]>; + +// Population Count +def VPOPCNTB : VXForm_2<1795, (outs vrrc:$vD), (ins vrrc:$vB), + "vpopcntb $vD, $vB", IIC_VecGeneral, + [(set v16i8:$vD, (ctpop v16i8:$vB))]>; +def VPOPCNTH : VXForm_2<1859, (outs vrrc:$vD), (ins vrrc:$vB), + "vpopcnth $vD, $vB", IIC_VecGeneral, + [(set v8i16:$vD, (ctpop v8i16:$vB))]>; +def VPOPCNTW : VXForm_2<1923, (outs vrrc:$vD), (ins vrrc:$vB), + "vpopcntw $vD, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (ctpop v4i32:$vB))]>; +def VPOPCNTD : VXForm_2<1987, (outs vrrc:$vD), (ins vrrc:$vB), + "vpopcntd $vD, $vB", IIC_VecGeneral, + [(set v2i64:$vD, (ctpop v2i64:$vB))]>; + +let isCommutable = 1 in { +// FIXME: Use AddedComplexity > 400 to ensure these patterns match before the +// VSX equivalents. We need to fix this up at some point. Two possible +// solutions for this problem: +// 1. Disable Altivec patterns that compete with VSX patterns using the +// !HasVSX predicate. This essentially favours VSX over Altivec, in +// hopes of reducing register pressure (larger register set using VSX +// instructions than VMX instructions) +// 2. Employ a more disciplined use of AddedComplexity, which would provide +// more fine-grained control than option 1. This would be beneficial +// if we find situations where Altivec is really preferred over VSX. +def VEQV : VXForm_1<1668, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "veqv $vD, $vA, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (vnot_ppc (xor v4i32:$vA, v4i32:$vB)))]>; +def VNAND : VXForm_1<1412, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vnand $vD, $vA, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (vnot_ppc (and v4i32:$vA, v4i32:$vB)))]>; +} // isCommutable + +def VORC : VXForm_1<1348, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + "vorc $vD, $vA, $vB", IIC_VecGeneral, + [(set v4i32:$vD, (or v4i32:$vA, + (vnot_ppc v4i32:$vB)))]>; + +// i64 element comparisons. +def VCMPEQUD : VCMP <199, "vcmpequd $vD, $vA, $vB" , v2i64>; +def VCMPEQUDo : VCMPo<199, "vcmpequd. $vD, $vA, $vB", v2i64>; +def VCMPGTSD : VCMP <967, "vcmpgtsd $vD, $vA, $vB" , v2i64>; +def VCMPGTSDo : VCMPo<967, "vcmpgtsd. $vD, $vA, $vB", v2i64>; +def VCMPGTUD : VCMP <711, "vcmpgtud $vD, $vA, $vB" , v2i64>; +def VCMPGTUDo : VCMPo<711, "vcmpgtud. $vD, $vA, $vB", v2i64>; + +// The cryptography instructions that do not require Category:Vector.Crypto +def VPMSUMB : VX1_Int_Ty<1032, "vpmsumb", + int_ppc_altivec_crypto_vpmsumb, v16i8>; +def VPMSUMH : VX1_Int_Ty<1096, "vpmsumh", + int_ppc_altivec_crypto_vpmsumh, v8i16>; +def VPMSUMW : VX1_Int_Ty<1160, "vpmsumw", + int_ppc_altivec_crypto_vpmsumw, v4i32>; +def VPMSUMD : VX1_Int_Ty<1224, "vpmsumd", + int_ppc_altivec_crypto_vpmsumd, v2i64>; +def VPERMXOR : VA1a_Int_Ty<45, "vpermxor", + int_ppc_altivec_crypto_vpermxor, v16i8>; + +} // end HasP8Altivec + +// Crypto instructions (from builtins) +let Predicates = [HasP8Crypto] in { +def VSHASIGMAW : VXCR_Int_Ty<1666, "vshasigmaw", + int_ppc_altivec_crypto_vshasigmaw, v4i32>; +def VSHASIGMAD : VXCR_Int_Ty<1730, "vshasigmad", + int_ppc_altivec_crypto_vshasigmad, v2i64>; +def VCIPHER : VX1_Int_Ty<1288, "vcipher", int_ppc_altivec_crypto_vcipher, + v2i64>; +def VCIPHERLAST : VX1_Int_Ty<1289, "vcipherlast", + int_ppc_altivec_crypto_vcipherlast, v2i64>; +def VNCIPHER : VX1_Int_Ty<1352, "vncipher", + int_ppc_altivec_crypto_vncipher, v2i64>; +def VNCIPHERLAST : VX1_Int_Ty<1353, "vncipherlast", + int_ppc_altivec_crypto_vncipherlast, v2i64>; +def VSBOX : VXBX_Int_Ty<1480, "vsbox", int_ppc_altivec_crypto_vsbox, v2i64>; +} // HasP8Crypto