//
//===----------------------------------------------------------------------===//
+// *********************************** NOTE ***********************************
+// ** For POWER8 Little Endian, the VSX swap optimization relies on knowing **
+// ** which VMX and VSX instructions are lane-sensitive and which are not. **
+// ** A lane-sensitive instruction relies, implicitly or explicitly, on **
+// ** whether lanes are numbered from left to right. An instruction like **
+// ** VADDFP is not lane-sensitive, because each lane of the result vector **
+// ** relies only on the corresponding lane of the source vectors. However, **
+// ** an instruction like VMULESB is lane-sensitive, because "even" and **
+// ** "odd" lanes are different for big-endian and little-endian numbering. **
+// ** **
+// ** When adding new VMX and VSX instructions, please consider whether they **
+// ** are lane-sensitive. If so, they must be added to a switch statement **
+// ** in PPCVSXSwapRemoval::gatherVectorInstructions(). **
+// ****************************************************************************
+
//===----------------------------------------------------------------------===//
// Altivec transformation functions and pattern fragments.
//
def vpkuhum_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVPKUHUMShuffleMask(cast<ShuffleVectorSDNode>(N), false,
- *CurDAG);
+ return PPC::isVPKUHUMShuffleMask(cast<ShuffleVectorSDNode>(N), 0, *CurDAG);
}]>;
def vpkuwum_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVPKUWUMShuffleMask(cast<ShuffleVectorSDNode>(N), false,
- *CurDAG);
+ return PPC::isVPKUWUMShuffleMask(cast<ShuffleVectorSDNode>(N), 0, *CurDAG);
+}]>;
+def vpkudum_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVPKUDUMShuffleMask(cast<ShuffleVectorSDNode>(N), 0, *CurDAG);
}]>;
def vpkuhum_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVPKUHUMShuffleMask(cast<ShuffleVectorSDNode>(N), true,
- *CurDAG);
+ return PPC::isVPKUHUMShuffleMask(cast<ShuffleVectorSDNode>(N), 1, *CurDAG);
}]>;
def vpkuwum_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVPKUWUMShuffleMask(cast<ShuffleVectorSDNode>(N), true,
- *CurDAG);
+ return PPC::isVPKUWUMShuffleMask(cast<ShuffleVectorSDNode>(N), 1, *CurDAG);
+}]>;
+def vpkudum_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVPKUDUMShuffleMask(cast<ShuffleVectorSDNode>(N), 1, *CurDAG);
}]>;
+// 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<ShuffleVectorSDNode>(N), 2, *CurDAG);
+}]>;
+def vpkuwum_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVPKUWUMShuffleMask(cast<ShuffleVectorSDNode>(N), 2, *CurDAG);
+}]>;
+def vpkudum_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVPKUDUMShuffleMask(cast<ShuffleVectorSDNode>(N), 2, *CurDAG);
+}]>;
def vmrglb_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
}]>;
+def vmrgew_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 0, *CurDAG);
+}]>;
+def vmrgow_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 0, *CurDAG);
+}]>;
+def vmrgew_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 1, *CurDAG);
+}]>;
+def vmrgow_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 1, *CurDAG);
+}]>;
+def vmrgew_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 2, *CurDAG);
+}]>;
+def vmrgow_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 2, *CurDAG);
+}]>;
+
+
+
def VSLDOI_get_imm : SDNodeXForm<vector_shuffle, [{
- return getI32Imm(PPC::isVSLDOIShuffleMask(N, false, *CurDAG));
+ return getI32Imm(PPC::isVSLDOIShuffleMask(N, 0, *CurDAG), SDLoc(N));
}]>;
def vsldoi_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVSLDOIShuffleMask(N, false, *CurDAG) != -1;
+ 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<vector_shuffle, [{
- return getI32Imm(PPC::isVSLDOIShuffleMask(N, true, *CurDAG));
+ return getI32Imm(PPC::isVSLDOIShuffleMask(N, 1, *CurDAG), SDLoc(N));
}]>;
def vsldoi_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
- return PPC::isVSLDOIShuffleMask(N, true, *CurDAG) != -1;
+ 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<vector_shuffle, [{
+ return getI32Imm(PPC::isVSLDOIShuffleMask(N, 2, *CurDAG), SDLoc(N));
+}]>;
+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<vector_shuffle, [{
- return getI32Imm(PPC::getVSPLTImmediate(N, 1, *CurDAG));
+ return getI32Imm(PPC::getVSPLTImmediate(N, 1, *CurDAG), SDLoc(N));
}]>;
def vspltb_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
return PPC::isSplatShuffleMask(cast<ShuffleVectorSDNode>(N), 1);
}], VSPLTB_get_imm>;
def VSPLTH_get_imm : SDNodeXForm<vector_shuffle, [{
- return getI32Imm(PPC::getVSPLTImmediate(N, 2, *CurDAG));
+ return getI32Imm(PPC::getVSPLTImmediate(N, 2, *CurDAG), SDLoc(N));
}]>;
def vsplth_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
return PPC::isSplatShuffleMask(cast<ShuffleVectorSDNode>(N), 2);
}], VSPLTH_get_imm>;
def VSPLTW_get_imm : SDNodeXForm<vector_shuffle, [{
- return getI32Imm(PPC::getVSPLTImmediate(N, 4, *CurDAG));
+ return getI32Imm(PPC::getVSPLTImmediate(N, 4, *CurDAG), SDLoc(N));
}]>;
def vspltw_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
(vector_shuffle node:$lhs, node:$rhs), [{
!strconcat(opc, " $vD, $vB"), IIC_VecFP,
[(set OutTy:$vD, (IntID InTy:$vB))]>;
+class VXBX_Int_Ty<bits<11> xo, string opc, Intrinsic IntID, ValueType Ty>
+ : VXForm_BX<xo, (outs vrrc:$vD), (ins vrrc:$vA),
+ !strconcat(opc, " $vD, $vA"), IIC_VecFP,
+ [(set Ty:$vD, (IntID Ty:$vA))]>;
+
+class VXCR_Int_Ty<bits<11> xo, string opc, Intrinsic IntID, ValueType Ty>
+ : VXForm_CR<xo, (outs vrrc:$vD), (ins vrrc:$vA, u1imm:$ST, u4imm:$SIX),
+ !strconcat(opc, " $vD, $vA, $ST, $SIX"), IIC_VecFP,
+ [(set Ty:$vD, (IntID Ty:$vA, imm:$ST, imm:$SIX))]>;
+
//===----------------------------------------------------------------------===//
// Instruction Definitions.
def HasAltivec : Predicate<"PPCSubTarget->hasAltivec()">;
let Predicates = [HasAltivec] in {
-let isCodeGenOnly = 1 in {
-def DSS : DSS_Form<822, (outs),
- (ins u5imm:$ZERO0, u5imm:$STRM,u5imm:$ZERO1,u5imm:$ZERO2),
- "dss $STRM", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
-def DSSALL : DSS_Form<822, (outs),
- (ins u5imm:$ONE, u5imm:$ZERO0,u5imm:$ZERO1,u5imm:$ZERO2),
- "dssall", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
-def DST : DSS_Form<342, (outs),
- (ins u5imm:$ZERO, u5imm:$STRM, gprc:$rA, gprc:$rB),
- "dst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
-def DSTT : DSS_Form<342, (outs),
- (ins u5imm:$ONE, u5imm:$STRM, gprc:$rA, gprc:$rB),
- "dstt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
-def DSTST : DSS_Form<374, (outs),
- (ins u5imm:$ZERO, u5imm:$STRM, gprc:$rA, gprc:$rB),
- "dstst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
-def DSTSTT : DSS_Form<374, (outs),
- (ins u5imm:$ONE, u5imm:$STRM, gprc:$rA, gprc:$rB),
- "dststt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
- Deprecated<DeprecatedDST>;
+def DSS : DSS_Form<0, 822, (outs), (ins u5imm:$STRM),
+ "dss $STRM", IIC_LdStLoad /*FIXME*/, [(int_ppc_altivec_dss imm:$STRM)]>,
+ Deprecated<DeprecatedDST> {
+ let A = 0;
+ let B = 0;
+}
+
+def DSSALL : DSS_Form<1, 822, (outs), (ins),
+ "dssall", IIC_LdStLoad /*FIXME*/, [(int_ppc_altivec_dssall)]>,
+ Deprecated<DeprecatedDST> {
+ let STRM = 0;
+ let A = 0;
+ let B = 0;
+}
-def DST64 : DSS_Form<342, (outs),
- (ins u5imm:$ZERO, u5imm:$STRM, g8rc:$rA, gprc:$rB),
- "dst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
+def DST : DSS_Form<0, 342, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB),
+ "dst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/,
+ [(int_ppc_altivec_dst i32:$rA, i32:$rB, imm:$STRM)]>,
Deprecated<DeprecatedDST>;
-def DSTT64 : DSS_Form<342, (outs),
- (ins u5imm:$ONE, u5imm:$STRM, g8rc:$rA, gprc:$rB),
- "dstt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
+
+def DSTT : DSS_Form<1, 342, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB),
+ "dstt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/,
+ [(int_ppc_altivec_dstt i32:$rA, i32:$rB, imm:$STRM)]>,
Deprecated<DeprecatedDST>;
-def DSTST64 : DSS_Form<374, (outs),
- (ins u5imm:$ZERO, u5imm:$STRM, g8rc:$rA, gprc:$rB),
- "dstst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
+
+def DSTST : DSS_Form<0, 374, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB),
+ "dstst $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/,
+ [(int_ppc_altivec_dstst i32:$rA, i32:$rB, imm:$STRM)]>,
Deprecated<DeprecatedDST>;
-def DSTSTT64 : DSS_Form<374, (outs),
- (ins u5imm:$ONE, u5imm:$STRM, g8rc:$rA, gprc:$rB),
- "dststt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/, []>,
+
+def DSTSTT : DSS_Form<1, 374, (outs), (ins u5imm:$STRM, gprc:$rA, gprc:$rB),
+ "dststt $rA, $rB, $STRM", IIC_LdStLoad /*FIXME*/,
+ [(int_ppc_altivec_dststt i32:$rA, i32:$rB, imm:$STRM)]>,
Deprecated<DeprecatedDST>;
+
+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<DeprecatedDST>;
+
+ 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<DeprecatedDST>;
+
+ 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<DeprecatedDST>;
+
+ 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<DeprecatedDST>;
}
def MFVSCR : VXForm_4<1540, (outs vrrc:$vD), (ins),
"mtvscr $vB", IIC_LdStLoad,
[(int_ppc_altivec_mtvscr v4i32:$vB)]>;
-let canFoldAsLoad = 1, PPC970_Unit = 2 in { // Loads.
+let 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:$src))]>;
def VPKSHUS : VX1_Int_Ty2<270, "vpkshus", int_ppc_altivec_vpkshus,
v16i8, v8i16>;
def VPKSWSS : VX1_Int_Ty2<462, "vpkswss", int_ppc_altivec_vpkswss,
- v16i8, v4i32>;
+ v8i16, v4i32>;
def VPKSWUS : VX1_Int_Ty2<334, "vpkswus", int_ppc_altivec_vpkswus,
v8i16, v4i32>;
def VPKUHUM : VXForm_1<14, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
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, $vB", v4i32>;
-
+
let isCodeGenOnly = 1 in {
def V_SET0B : VXForm_setzero<1220, (outs vrrc:$vD), (ins),
"vxor $vD, $vD, $vD", IIC_VecFP,
// Additional Altivec Patterns
//
-// DS* intrinsics
-def : Pat<(int_ppc_altivec_dssall), (DSSALL 1, 0, 0, 0)>;
-def : Pat<(int_ppc_altivec_dss imm:$STRM), (DSS 0, imm:$STRM, 0, 0)>;
-
-// * 32-bit
-def : Pat<(int_ppc_altivec_dst i32:$rA, i32:$rB, imm:$STRM),
- (DST 0, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dstt i32:$rA, i32:$rB, imm:$STRM),
- (DSTT 1, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dstst i32:$rA, i32:$rB, imm:$STRM),
- (DSTST 0, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dststt i32:$rA, i32:$rB, imm:$STRM),
- (DSTSTT 1, imm:$STRM, $rA, $rB)>;
-
-// * 64-bit
-def : Pat<(int_ppc_altivec_dst i64:$rA, i32:$rB, imm:$STRM),
- (DST64 0, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dstt i64:$rA, i32:$rB, imm:$STRM),
- (DSTT64 1, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dstst i64:$rA, i32:$rB, imm:$STRM),
- (DSTST64 0, imm:$STRM, $rA, $rB)>;
-def : Pat<(int_ppc_altivec_dststt i64:$rA, i32:$rB, imm:$STRM),
- (DSTSTT64 1, imm:$STRM, $rA, $rB)>;
-
// Loads.
def : Pat<(v4i32 (load xoaddr:$src)), (LVX xoaddr:$src)>;
def : Pat<(v16i8 (bitconvert (v8i16 VRRC:$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:$src)>;
+def : Pat<(v16i8 (bitconvert (v1i128 VRRC:$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:$src)>;
+def : Pat<(v8i16 (bitconvert (v1i128 VRRC:$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:$src)>;
+def : Pat<(v4i32 (bitconvert (v1i128 VRRC:$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:$src)>;
+def : Pat<(v4f32 (bitconvert (v1i128 VRRC:$src))), (v4f32 VRRC:$src)>;
+
+def : Pat<(v2i64 (bitconvert (v16i8 VRRC:$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:$src)>;
+def : Pat<(v2i64 (bitconvert (v1i128 VRRC:$src))), (v2i64 VRRC:$src)>;
+
+def : Pat<(v1i128 (bitconvert (v16i8 VRRC:$src))), (v1i128 VRRC:$src)>;
+def : Pat<(v1i128 (bitconvert (v8i16 VRRC:$src))), (v1i128 VRRC:$src)>;
+def : Pat<(v1i128 (bitconvert (v4i32 VRRC:$src))), (v1i128 VRRC:$src)>;
+def : Pat<(v1i128 (bitconvert (v4f32 VRRC:$src))), (v1i128 VRRC:$src)>;
+def : Pat<(v1i128 (bitconvert (v2i64 VRRC:$src))), (v1i128 VRRC:$src)>;
// Shuffles.
def:Pat<(vpkuhum_unary_shuffle v16i8:$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, $vA)>;
} // end HasAltivec
+def HasP8Altivec : Predicate<"PPCSubTarget->hasP8Altivec()">;
+def HasP8Crypto : Predicate<"PPCSubTarget->hasP8Crypto()">;
+let Predicates = [HasP8Altivec] in {
+
+let isCommutable = 1 in {
+def VMULESW : VX1_Int_Ty2<904, "vmulesw", int_ppc_altivec_vmulesw,
+ v2i64, v4i32>;
+def VMULEUW : VX1_Int_Ty2<648, "vmuleuw", int_ppc_altivec_vmuleuw,
+ v2i64, v4i32>;
+def VMULOSW : VX1_Int_Ty2<392, "vmulosw", int_ppc_altivec_vmulosw,
+ v2i64, v4i32>;
+def VMULOUW : VX1_Int_Ty2<136, "vmulouw", int_ppc_altivec_vmulouw,
+ v2i64, v4i32>;
+def VMULUWM : VXForm_1<137, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vmuluwm $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v4i32:$vD, (mul v4i32:$vA, v4i32:$vB))]>;
+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 VMINUD : VX1_Int_Ty<706, "vminud", int_ppc_altivec_vminud, v2i64>;
+} // isCommutable
+
+// Vector merge
+def VMRGEW : VXForm_1<1932, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vmrgew $vD, $vA, $vB", IIC_VecFP,
+ [(set v16i8:$vD, (vmrgew_shuffle v16i8:$vA, v16i8:$vB))]>;
+def VMRGOW : VXForm_1<1676, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vmrgow $vD, $vA, $vB", IIC_VecFP,
+ [(set v16i8:$vD, (vmrgow_shuffle v16i8:$vA, v16i8:$vB))]>;
+
+// Match vmrgew(x,x) and vmrgow(x,x)
+def:Pat<(vmrgew_unary_shuffle v16i8:$vA, undef),
+ (VMRGEW $vA, $vA)>;
+def:Pat<(vmrgow_unary_shuffle v16i8:$vA, undef),
+ (VMRGOW $vA, $vA)>;
+
+// Match vmrgew(y,x) and vmrgow(y,x), i.e., swapped operands. These fragments
+// are matched for little-endian, where the inputs must be swapped for correct
+// semantics.w
+def:Pat<(vmrgew_swapped_shuffle v16i8:$vA, v16i8:$vB),
+ (VMRGEW $vB, $vA)>;
+def:Pat<(vmrgow_swapped_shuffle v16i8:$vA, v16i8:$vB),
+ (VMRGOW $vB, $vA)>;
+
+
+// Vector shifts
+def VRLD : VX1_Int_Ty<196, "vrld", int_ppc_altivec_vrld, v2i64>;
+def VSLD : VXForm_1<1476, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vsld $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v2i64:$vD, (shl v2i64:$vA, v2i64:$vB))]>;
+def VSRD : VXForm_1<1732, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vsrd $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v2i64:$vD, (srl v2i64:$vA, v2i64:$vB))]>;
+def VSRAD : VXForm_1<964, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vsrad $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v2i64:$vD, (sra v2i64:$vA, v2i64:$vB))]>;
+
+// 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))]>;
+def VADDUQM : VXForm_1<256, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vadduqm $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v1i128:$vD, (add v1i128:$vA, v1i128:$vB))]>;
+} // isCommutable
+
+// Vector Quadword Add
+def VADDEUQM : VA1a_Int_Ty<60, "vaddeuqm", int_ppc_altivec_vaddeuqm, v1i128>;
+def VADDCUQ : VX1_Int_Ty<320, "vaddcuq", int_ppc_altivec_vaddcuq, v1i128>;
+def VADDECUQ : VA1a_Int_Ty<61, "vaddecuq", int_ppc_altivec_vaddecuq, v1i128>;
+
+// Vector Doubleword Subtract
+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))]>;
+
+// Vector Quadword Subtract
+def VSUBUQM : VXForm_1<1280, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vsubuqm $vD, $vA, $vB", IIC_VecGeneral,
+ [(set v1i128:$vD, (sub v1i128:$vA, v1i128:$vB))]>;
+def VSUBEUQM : VA1a_Int_Ty<62, "vsubeuqm", int_ppc_altivec_vsubeuqm, v1i128>;
+def VSUBCUQ : VX1_Int_Ty<1344, "vsubcuq", int_ppc_altivec_vsubcuq, v1i128>;
+def VSUBECUQ : VA1a_Int_Ty<63, "vsubecuq", int_ppc_altivec_vsubecuq, v1i128>;
+
+// 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, $vB", v2i64>;
+
+// The cryptography instructions that do not require Category:Vector.Crypto
+def VPMSUMB : VX1_Int_Ty<1032, "vpmsumb",
+ int_ppc_altivec_crypto_vpmsumb, v16i8>;
+def VPMSUMH : VX1_Int_Ty<1096, "vpmsumh",
+ int_ppc_altivec_crypto_vpmsumh, v8i16>;
+def VPMSUMW : VX1_Int_Ty<1160, "vpmsumw",
+ int_ppc_altivec_crypto_vpmsumw, v4i32>;
+def VPMSUMD : VX1_Int_Ty<1224, "vpmsumd",
+ int_ppc_altivec_crypto_vpmsumd, v2i64>;
+def VPERMXOR : VA1a_Int_Ty<45, "vpermxor",
+ int_ppc_altivec_crypto_vpermxor, v16i8>;
+
+// Vector doubleword integer pack and unpack.
+def VPKSDSS : VX1_Int_Ty2<1486, "vpksdss", int_ppc_altivec_vpksdss,
+ v4i32, v2i64>;
+def VPKSDUS : VX1_Int_Ty2<1358, "vpksdus", int_ppc_altivec_vpksdus,
+ v4i32, v2i64>;
+def VPKUDUM : VXForm_1<1102, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vpkudum $vD, $vA, $vB", IIC_VecFP,
+ [(set v16i8:$vD,
+ (vpkudum_shuffle v16i8:$vA, v16i8:$vB))]>;
+def VPKUDUS : VX1_Int_Ty2<1230, "vpkudus", int_ppc_altivec_vpkudus,
+ v4i32, v2i64>;
+def VUPKHSW : VX2_Int_Ty2<1614, "vupkhsw", int_ppc_altivec_vupkhsw,
+ v2i64, v4i32>;
+def VUPKLSW : VX2_Int_Ty2<1742, "vupklsw", int_ppc_altivec_vupklsw,
+ v2i64, v4i32>;
+
+// Shuffle patterns for unary and swapped (LE) vector pack modulo.
+def:Pat<(vpkudum_unary_shuffle v16i8:$vA, undef),
+ (VPKUDUM $vA, $vA)>;
+def:Pat<(vpkudum_swapped_shuffle v16i8:$vA, v16i8:$vB),
+ (VPKUDUM $vB, $vA)>;
+
+def VGBBD : VX2_Int_Ty2<1292, "vgbbd", int_ppc_altivec_vgbbd, v16i8, v16i8>;
+def VBPERMQ : VX1_Int_Ty2<1356, "vbpermq", int_ppc_altivec_vbpermq,
+ v2i64, 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
projects / oota-llvm.git / blobdiff