aboutsummaryrefslogtreecommitdiff
path: root/lib/Target/PowerPC/PPCInstrAltivec.td
diff options
context:
space:
mode:
authorAlexander Kornienko <alexfh@google.com>2013-04-03 14:07:16 +0000
committerAlexander Kornienko <alexfh@google.com>2013-04-03 14:07:16 +0000
commite133bc868944822bf8961f825d3aa63d6fa48fb7 (patch)
treeebbd4a8040181471467a9737d90d94dc6b58b316 /lib/Target/PowerPC/PPCInstrAltivec.td
parent647735c781c5b37061ee03d6e9e6c7dda92218e2 (diff)
parent080e3c523e87ec68ca1ea5db4cd49816028dd8bd (diff)
Updating branches/google/stable to r178511stable
git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/google/stable@178655 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/PowerPC/PPCInstrAltivec.td')
-rw-r--r--lib/Target/PowerPC/PPCInstrAltivec.td578
1 files changed, 342 insertions, 236 deletions
diff --git a/lib/Target/PowerPC/PPCInstrAltivec.td b/lib/Target/PowerPC/PPCInstrAltivec.td
index 0cf28ae4b5..fff91df418 100644
--- a/lib/Target/PowerPC/PPCInstrAltivec.td
+++ b/lib/Target/PowerPC/PPCInstrAltivec.td
@@ -161,27 +161,90 @@ def vecspltisw : PatLeaf<(build_vector), [{
//===----------------------------------------------------------------------===//
// Helpers for defining instructions that directly correspond to intrinsics.
-// VA1a_Int - A VAForm_1a intrinsic definition.
+// VA1a_Int - A VAForm_1a intrinsic definition of generic type.
class VA1a_Int<bits<6> xo, string opc, Intrinsic IntID>
: VAForm_1a<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, VRRC:$vC),
!strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP,
[(set VRRC:$vD, (IntID VRRC:$vA, VRRC:$vB, VRRC:$vC))]>;
-// VX1_Int - A VXForm_1 intrinsic definition.
+// VA1a_Int_Ty - A VAForm_1a intrinsic definition of specific type.
+class VA1a_Int_Ty<bits<6> xo, string opc, Intrinsic IntID, ValueType Ty>
+ : VAForm_1a<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, VRRC:$vC),
+ !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP,
+ [(set Ty:$vD, (IntID Ty:$vA, Ty:$vB, Ty:$vC))]>;
+
+// VA1a_Int_Ty2 - A VAForm_1a intrinsic definition where the type of the
+// inputs doesn't match the type of the output.
+class VA1a_Int_Ty2<bits<6> xo, string opc, Intrinsic IntID, ValueType OutTy,
+ ValueType InTy>
+ : VAForm_1a<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, VRRC:$vC),
+ !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP,
+ [(set OutTy:$vD, (IntID InTy:$vA, InTy:$vB, InTy:$vC))]>;
+
+// VA1a_Int_Ty3 - A VAForm_1a intrinsic definition where there are two
+// input types and an output type.
+class VA1a_Int_Ty3<bits<6> xo, string opc, Intrinsic IntID, ValueType OutTy,
+ ValueType In1Ty, ValueType In2Ty>
+ : VAForm_1a<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, VRRC:$vC),
+ !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP,
+ [(set OutTy:$vD,
+ (IntID In1Ty:$vA, In1Ty:$vB, In2Ty:$vC))]>;
+
+// VX1_Int - A VXForm_1 intrinsic definition of generic type.
class VX1_Int<bits<11> xo, string opc, Intrinsic IntID>
: VXForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
!strconcat(opc, " $vD, $vA, $vB"), VecFP,
[(set VRRC:$vD, (IntID VRRC:$vA, VRRC:$vB))]>;
-// VX2_Int - A VXForm_2 intrinsic definition.
+// VX1_Int_Ty - A VXForm_1 intrinsic definition of specific type.
+class VX1_Int_Ty<bits<11> xo, string opc, Intrinsic IntID, ValueType Ty>
+ : VXForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
+ !strconcat(opc, " $vD, $vA, $vB"), VecFP,
+ [(set Ty:$vD, (IntID Ty:$vA, Ty:$vB))]>;
+
+// 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<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy,
+ ValueType InTy>
+ : VXForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
+ !strconcat(opc, " $vD, $vA, $vB"), VecFP,
+ [(set OutTy:$vD, (IntID InTy:$vA, InTy:$vB))]>;
+
+// VX1_Int_Ty3 - A VXForm_1 intrinsic definition where there are two
+// input types and an output type.
+class VX1_Int_Ty3<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy,
+ ValueType In1Ty, ValueType In2Ty>
+ : VXForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
+ !strconcat(opc, " $vD, $vA, $vB"), VecFP,
+ [(set OutTy:$vD, (IntID In1Ty:$vA, In2Ty:$vB))]>;
+
+// VX2_Int - A VXForm_1 intrinsic definition of generic type.
class VX2_Int<bits<11> xo, string opc, Intrinsic IntID>
: VXForm_2<xo, (outs VRRC:$vD), (ins VRRC:$vB),
!strconcat(opc, " $vD, $vB"), VecFP,
[(set VRRC:$vD, (IntID VRRC:$vB))]>;
+// VX2_Int_SP - A VXForm_2 intrinsic definition of vector single-precision type.
+class VX2_Int_SP<bits<11> xo, string opc, Intrinsic IntID>
+ : VXForm_2<xo, (outs VRRC:$vD), (ins VRRC:$vB),
+ !strconcat(opc, " $vD, $vB"), VecFP,
+ [(set v4f32:$vD, (IntID v4f32:$vB))]>;
+
+// VX2_Int_Ty2 - A VXForm_2 intrinsic definition where the type of the
+// inputs doesn't match the type of the output.
+class VX2_Int_Ty2<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy,
+ ValueType InTy>
+ : VXForm_2<xo, (outs VRRC:$vD), (ins VRRC:$vB),
+ !strconcat(opc, " $vD, $vB"), VecFP,
+ [(set OutTy:$vD, (IntID InTy:$vB))]>;
+
//===----------------------------------------------------------------------===//
// Instruction Definitions.
+def HasAltivec : Predicate<"PPCSubTarget.hasAltivec()">;
+let Predicates = [HasAltivec] in {
+
+let isCodeGenOnly = 1 in {
def DSS : DSS_Form<822, (outs),
(ins u5imm:$ZERO0, u5imm:$STRM,u5imm:$ZERO1,u5imm:$ZERO2),
"dss $STRM", LdStLoad /*FIXME*/, []>;
@@ -213,74 +276,79 @@ def DSTST64 : DSS_Form<374, (outs),
def DSTSTT64 : DSS_Form<374, (outs),
(ins u5imm:$ONE, u5imm:$STRM, G8RC:$rA, GPRC:$rB),
"dststt $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>;
+}
def MFVSCR : VXForm_4<1540, (outs VRRC:$vD), (ins),
"mfvscr $vD", LdStStore,
- [(set VRRC:$vD, (int_ppc_altivec_mfvscr))]>;
+ [(set v8i16:$vD, (int_ppc_altivec_mfvscr))]>;
def MTVSCR : VXForm_5<1604, (outs), (ins VRRC:$vB),
"mtvscr $vB", LdStLoad,
- [(int_ppc_altivec_mtvscr VRRC:$vB)]>;
+ [(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, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>;
+ [(set v16i8:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>;
def LVEHX: XForm_1<31, 39, (outs VRRC:$vD), (ins memrr:$src),
"lvehx $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>;
+ [(set v8i16:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>;
def LVEWX: XForm_1<31, 71, (outs VRRC:$vD), (ins memrr:$src),
"lvewx $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>;
+ [(set v4i32:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>;
def LVX : XForm_1<31, 103, (outs VRRC:$vD), (ins memrr:$src),
"lvx $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>;
+ [(set v4i32:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>;
def LVXL : XForm_1<31, 359, (outs VRRC:$vD), (ins memrr:$src),
"lvxl $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>;
+ [(set v4i32:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>;
}
def LVSL : XForm_1<31, 6, (outs VRRC:$vD), (ins memrr:$src),
"lvsl $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>,
+ [(set v16i8:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>,
PPC970_Unit_LSU;
def LVSR : XForm_1<31, 38, (outs VRRC:$vD), (ins memrr:$src),
"lvsr $vD, $src", LdStLoad,
- [(set VRRC:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>,
+ [(set v16i8:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>,
PPC970_Unit_LSU;
let PPC970_Unit = 2 in { // Stores.
def STVEBX: XForm_8<31, 135, (outs), (ins VRRC:$rS, memrr:$dst),
"stvebx $rS, $dst", LdStStore,
- [(int_ppc_altivec_stvebx VRRC:$rS, xoaddr:$dst)]>;
+ [(int_ppc_altivec_stvebx v16i8:$rS, xoaddr:$dst)]>;
def STVEHX: XForm_8<31, 167, (outs), (ins VRRC:$rS, memrr:$dst),
"stvehx $rS, $dst", LdStStore,
- [(int_ppc_altivec_stvehx VRRC:$rS, xoaddr:$dst)]>;
+ [(int_ppc_altivec_stvehx v8i16:$rS, xoaddr:$dst)]>;
def STVEWX: XForm_8<31, 199, (outs), (ins VRRC:$rS, memrr:$dst),
"stvewx $rS, $dst", LdStStore,
- [(int_ppc_altivec_stvewx VRRC:$rS, xoaddr:$dst)]>;
+ [(int_ppc_altivec_stvewx v4i32:$rS, xoaddr:$dst)]>;
def STVX : XForm_8<31, 231, (outs), (ins VRRC:$rS, memrr:$dst),
"stvx $rS, $dst", LdStStore,
- [(int_ppc_altivec_stvx VRRC:$rS, xoaddr:$dst)]>;
+ [(int_ppc_altivec_stvx v4i32:$rS, xoaddr:$dst)]>;
def STVXL : XForm_8<31, 487, (outs), (ins VRRC:$rS, memrr:$dst),
"stvxl $rS, $dst", LdStStore,
- [(int_ppc_altivec_stvxl VRRC:$rS, xoaddr:$dst)]>;
+ [(int_ppc_altivec_stvxl v4i32:$rS, xoaddr:$dst)]>;
}
let PPC970_Unit = 5 in { // VALU Operations.
// VA-Form instructions. 3-input AltiVec ops.
def VMADDFP : VAForm_1<46, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vC, VRRC:$vB),
"vmaddfp $vD, $vA, $vC, $vB", VecFP,
- [(set VRRC:$vD, (fma VRRC:$vA, VRRC:$vC, VRRC:$vB))]>;
+ [(set v4f32:$vD,
+ (fma v4f32:$vA, v4f32:$vC, v4f32:$vB))]>;
def VNMSUBFP: VAForm_1<47, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vC, VRRC:$vB),
"vnmsubfp $vD, $vA, $vC, $vB", VecFP,
- [(set VRRC:$vD, (fneg (fma VRRC:$vA, VRRC:$vC,
- (fneg VRRC:$vB))))]>;
+ [(set v4f32:$vD, (fneg (fma v4f32:$vA, v4f32:$vC,
+ (fneg v4f32:$vB))))]>;
+
+def VMHADDSHS : VA1a_Int_Ty<32, "vmhaddshs", int_ppc_altivec_vmhaddshs, v8i16>;
+def VMHRADDSHS : VA1a_Int_Ty<33, "vmhraddshs", int_ppc_altivec_vmhraddshs,
+ v8i16>;
+def VMLADDUHM : VA1a_Int_Ty<34, "vmladduhm", int_ppc_altivec_vmladduhm, v8i16>;
-def VMHADDSHS : VA1a_Int<32, "vmhaddshs", int_ppc_altivec_vmhaddshs>;
-def VMHRADDSHS : VA1a_Int<33, "vmhraddshs", int_ppc_altivec_vmhraddshs>;
-def VMLADDUHM : VA1a_Int<34, "vmladduhm", int_ppc_altivec_vmladduhm>;
-def VPERM : VA1a_Int<43, "vperm", int_ppc_altivec_vperm>;
-def VSEL : VA1a_Int<42, "vsel", int_ppc_altivec_vsel>;
+def VPERM : VA1a_Int_Ty3<43, "vperm", int_ppc_altivec_vperm,
+ v4i32, v4i32, v16i8>;
+def VSEL : VA1a_Int_Ty<42, "vsel", int_ppc_altivec_vsel, v4i32>;
// Shuffles.
def VSLDOI : VAForm_2<44, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, u5imm:$SH),
@@ -291,25 +359,25 @@ def VSLDOI : VAForm_2<44, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, u5imm:$SH),
// VX-Form instructions. AltiVec arithmetic ops.
def VADDFP : VXForm_1<10, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vaddfp $vD, $vA, $vB", VecFP,
- [(set VRRC:$vD, (fadd VRRC:$vA, VRRC:$vB))]>;
+ [(set v4f32:$vD, (fadd v4f32:$vA, v4f32:$vB))]>;
def VADDUBM : VXForm_1<0, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vaddubm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (add (v16i8 VRRC:$vA), VRRC:$vB))]>;
+ [(set v16i8:$vD, (add v16i8:$vA, v16i8:$vB))]>;
def VADDUHM : VXForm_1<64, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vadduhm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (add (v8i16 VRRC:$vA), VRRC:$vB))]>;
+ [(set v8i16:$vD, (add v8i16:$vA, v8i16:$vB))]>;
def VADDUWM : VXForm_1<128, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vadduwm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (add (v4i32 VRRC:$vA), VRRC:$vB))]>;
+ [(set v4i32:$vD, (add v4i32:$vA, v4i32:$vB))]>;
-def VADDCUW : VX1_Int<384, "vaddcuw", int_ppc_altivec_vaddcuw>;
-def VADDSBS : VX1_Int<768, "vaddsbs", int_ppc_altivec_vaddsbs>;
-def VADDSHS : VX1_Int<832, "vaddshs", int_ppc_altivec_vaddshs>;
-def VADDSWS : VX1_Int<896, "vaddsws", int_ppc_altivec_vaddsws>;
-def VADDUBS : VX1_Int<512, "vaddubs", int_ppc_altivec_vaddubs>;
-def VADDUHS : VX1_Int<576, "vadduhs", int_ppc_altivec_vadduhs>;
-def VADDUWS : VX1_Int<640, "vadduws", int_ppc_altivec_vadduws>;
+def VADDCUW : VX1_Int_Ty<384, "vaddcuw", int_ppc_altivec_vaddcuw, v4i32>;
+def VADDSBS : VX1_Int_Ty<768, "vaddsbs", int_ppc_altivec_vaddsbs, v16i8>;
+def VADDSHS : VX1_Int_Ty<832, "vaddshs", int_ppc_altivec_vaddshs, v8i16>;
+def VADDSWS : VX1_Int_Ty<896, "vaddsws", int_ppc_altivec_vaddsws, v4i32>;
+def VADDUBS : VX1_Int_Ty<512, "vaddubs", int_ppc_altivec_vaddubs, v16i8>;
+def VADDUHS : VX1_Int_Ty<576, "vadduhs", int_ppc_altivec_vadduhs, v8i16>;
+def VADDUWS : VX1_Int_Ty<640, "vadduws", int_ppc_altivec_vadduws, v4i32>;
def VAND : VXForm_1<1028, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
@@ -322,20 +390,20 @@ def VANDC : VXForm_1<1092, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
def VCFSX : VXForm_1<842, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB),
"vcfsx $vD, $vB, $UIMM", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vcfsx VRRC:$vB, imm:$UIMM))]>;
+ [(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, $UIMM", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vcfux VRRC:$vB, imm:$UIMM))]>;
+ [(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, $UIMM", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vctsxs VRRC:$vB, imm:$UIMM))]>;
+ [(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, $UIMM", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vctuxs VRRC:$vB, imm:$UIMM))]>;
+ [(set v4i32:$vD,
+ (int_ppc_altivec_vctuxs v4f32:$vB, imm:$UIMM))]>;
// Defines with the UIM field set to 0 for floating-point
// to integer (fp_to_sint/fp_to_uint) conversions and integer
@@ -343,49 +411,49 @@ def VCTUXS : VXForm_1<906, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB),
let VA = 0 in {
def VCFSX_0 : VXForm_1<842, (outs VRRC:$vD), (ins VRRC:$vB),
"vcfsx $vD, $vB, 0", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vcfsx VRRC:$vB, 0))]>;
+ [(set v4f32:$vD,
+ (int_ppc_altivec_vcfsx v4i32:$vB, 0))]>;
def VCTUXS_0 : VXForm_1<906, (outs VRRC:$vD), (ins VRRC:$vB),
"vctuxs $vD, $vB, 0", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vctuxs VRRC:$vB, 0))]>;
+ [(set v4i32:$vD,
+ (int_ppc_altivec_vctuxs v4f32:$vB, 0))]>;
def VCFUX_0 : VXForm_1<778, (outs VRRC:$vD), (ins VRRC:$vB),
"vcfux $vD, $vB, 0", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vcfux VRRC:$vB, 0))]>;
+ [(set v4f32:$vD,
+ (int_ppc_altivec_vcfux v4i32:$vB, 0))]>;
def VCTSXS_0 : VXForm_1<970, (outs VRRC:$vD), (ins VRRC:$vB),
"vctsxs $vD, $vB, 0", VecFP,
- [(set VRRC:$vD,
- (int_ppc_altivec_vctsxs VRRC:$vB, 0))]>;
+ [(set v4i32:$vD,
+ (int_ppc_altivec_vctsxs v4f32:$vB, 0))]>;
}
-def VEXPTEFP : VX2_Int<394, "vexptefp", int_ppc_altivec_vexptefp>;
-def VLOGEFP : VX2_Int<458, "vlogefp", int_ppc_altivec_vlogefp>;
-
-def VAVGSB : VX1_Int<1282, "vavgsb", int_ppc_altivec_vavgsb>;
-def VAVGSH : VX1_Int<1346, "vavgsh", int_ppc_altivec_vavgsh>;
-def VAVGSW : VX1_Int<1410, "vavgsw", int_ppc_altivec_vavgsw>;
-def VAVGUB : VX1_Int<1026, "vavgub", int_ppc_altivec_vavgub>;
-def VAVGUH : VX1_Int<1090, "vavguh", int_ppc_altivec_vavguh>;
-def VAVGUW : VX1_Int<1154, "vavguw", int_ppc_altivec_vavguw>;
-
-def VMAXFP : VX1_Int<1034, "vmaxfp", int_ppc_altivec_vmaxfp>;
-def VMAXSB : VX1_Int< 258, "vmaxsb", int_ppc_altivec_vmaxsb>;
-def VMAXSH : VX1_Int< 322, "vmaxsh", int_ppc_altivec_vmaxsh>;
-def VMAXSW : VX1_Int< 386, "vmaxsw", int_ppc_altivec_vmaxsw>;
-def VMAXUB : VX1_Int< 2, "vmaxub", int_ppc_altivec_vmaxub>;
-def VMAXUH : VX1_Int< 66, "vmaxuh", int_ppc_altivec_vmaxuh>;
-def VMAXUW : VX1_Int< 130, "vmaxuw", int_ppc_altivec_vmaxuw>;
-def VMINFP : VX1_Int<1098, "vminfp", int_ppc_altivec_vminfp>;
-def VMINSB : VX1_Int< 770, "vminsb", int_ppc_altivec_vminsb>;
-def VMINSH : VX1_Int< 834, "vminsh", int_ppc_altivec_vminsh>;
-def VMINSW : VX1_Int< 898, "vminsw", int_ppc_altivec_vminsw>;
-def VMINUB : VX1_Int< 514, "vminub", int_ppc_altivec_vminub>;
-def VMINUH : VX1_Int< 578, "vminuh", int_ppc_altivec_vminuh>;
-def VMINUW : VX1_Int< 642, "vminuw", int_ppc_altivec_vminuw>;
+def VEXPTEFP : VX2_Int_SP<394, "vexptefp", int_ppc_altivec_vexptefp>;
+def VLOGEFP : VX2_Int_SP<458, "vlogefp", int_ppc_altivec_vlogefp>;
+
+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>;
def VMRGHB : VXForm_1< 12, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vmrghb $vD, $vA, $vB", VecFP,
- [(set VRRC:$vD, (vmrghb_shuffle VRRC:$vA, VRRC:$vB))]>;
+ [(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, $vB", VecFP,
[(set VRRC:$vD, (vmrghh_shuffle VRRC:$vA, VRRC:$vB))]>;
@@ -394,7 +462,7 @@ def VMRGHW : VXForm_1<140, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
[(set VRRC:$vD, (vmrghw_shuffle VRRC:$vA, VRRC:$vB))]>;
def VMRGLB : VXForm_1<268, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vmrglb $vD, $vA, $vB", VecFP,
- [(set VRRC:$vD, (vmrglb_shuffle VRRC:$vA, VRRC:$vB))]>;
+ [(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, $vB", VecFP,
[(set VRRC:$vD, (vmrglh_shuffle VRRC:$vA, VRRC:$vB))]>;
@@ -402,55 +470,74 @@ def VMRGLW : VXForm_1<396, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vmrglw $vD, $vA, $vB", VecFP,
[(set VRRC:$vD, (vmrglw_shuffle VRRC:$vA, VRRC:$vB))]>;
-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", int_ppc_altivec_vmulesb>;
-def VMULESH : VX1_Int<840, "vmulesh", int_ppc_altivec_vmulesh>;
-def VMULEUB : VX1_Int<520, "vmuleub", int_ppc_altivec_vmuleub>;
-def VMULEUH : VX1_Int<584, "vmuleuh", int_ppc_altivec_vmuleuh>;
-def VMULOSB : VX1_Int<264, "vmulosb", int_ppc_altivec_vmulosb>;
-def VMULOSH : VX1_Int<328, "vmulosh", int_ppc_altivec_vmulosh>;
-def VMULOUB : VX1_Int< 8, "vmuloub", int_ppc_altivec_vmuloub>;
-def VMULOUH : VX1_Int< 72, "vmulouh", int_ppc_altivec_vmulouh>;
+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>;
+
+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>;
-def VREFP : VX2_Int<266, "vrefp", int_ppc_altivec_vrefp>;
-def VRFIM : VX2_Int<714, "vrfim", int_ppc_altivec_vrfim>;
-def VRFIN : VX2_Int<522, "vrfin", int_ppc_altivec_vrfin>;
-def VRFIP : VX2_Int<650, "vrfip", int_ppc_altivec_vrfip>;
-def VRFIZ : VX2_Int<586, "vrfiz", int_ppc_altivec_vrfiz>;
-def VRSQRTEFP : VX2_Int<330, "vrsqrtefp", int_ppc_altivec_vrsqrtefp>;
+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<74, "vsubcuw", int_ppc_altivec_vsubcuw>;
+def VSUBCUW : VX1_Int_Ty<74, "vsubcuw", int_ppc_altivec_vsubcuw, v4i32>;
def VSUBFP : VXForm_1<74, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vsubfp $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (fsub VRRC:$vA, VRRC:$vB))]>;
+ [(set v4f32:$vD, (fsub v4f32:$vA, v4f32:$vB))]>;
def VSUBUBM : VXForm_1<1024, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vsububm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (sub (v16i8 VRRC:$vA), VRRC:$vB))]>;
+ [(set v16i8:$vD, (sub v16i8:$vA, v16i8:$vB))]>;
def VSUBUHM : VXForm_1<1088, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vsubuhm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (sub (v8i16 VRRC:$vA), VRRC:$vB))]>;
+ [(set v8i16:$vD, (sub v8i16:$vA, v8i16:$vB))]>;
def VSUBUWM : VXForm_1<1152, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vsubuwm $vD, $vA, $vB", VecGeneral,
- [(set VRRC:$vD, (sub (v4i32 VRRC:$vA), VRRC:$vB))]>;
+ [(set v4i32:$vD, (sub v4i32:$vA, v4i32:$vB))]>;
-def VSUBSBS : VX1_Int<1792, "vsubsbs" , int_ppc_altivec_vsubsbs>;
-def VSUBSHS : VX1_Int<1856, "vsubshs" , int_ppc_altivec_vsubshs>;
-def VSUBSWS : VX1_Int<1920, "vsubsws" , int_ppc_altivec_vsubsws>;
-def VSUBUBS : VX1_Int<1536, "vsububs" , int_ppc_altivec_vsububs>;
-def VSUBUHS : VX1_Int<1600, "vsubuhs" , int_ppc_altivec_vsubuhs>;
-def VSUBUWS : VX1_Int<1664, "vsubuws" , int_ppc_altivec_vsubuws>;
-def VSUMSWS : VX1_Int<1928, "vsumsws" , int_ppc_altivec_vsumsws>;
-def VSUM2SWS: VX1_Int<1672, "vsum2sws", int_ppc_altivec_vsum2sws>;
-def VSUM4SBS: VX1_Int<1672, "vsum4sbs", int_ppc_altivec_vsum4sbs>;
-def VSUM4SHS: VX1_Int<1608, "vsum4shs", int_ppc_altivec_vsum4shs>;
-def VSUM4UBS: VX1_Int<1544, "vsum4ubs", int_ppc_altivec_vsum4ubs>;
+def VSUBSBS : VX1_Int_Ty<1792, "vsubsbs" , int_ppc_altivec_vsubsbs, v16i8>;
+def VSUBSHS : VX1_Int_Ty<1856, "vsubshs" , int_ppc_altivec_vsubshs, v8i16>;
+def VSUBSWS : VX1_Int_Ty<1920, "vsubsws" , int_ppc_altivec_vsubsws, v4i32>;
+def VSUBUBS : VX1_Int_Ty<1536, "vsububs" , int_ppc_altivec_vsububs, v16i8>;
+def VSUBUHS : VX1_Int_Ty<1600, "vsubuhs" , int_ppc_altivec_vsubuhs, v8i16>;
+def VSUBUWS : VX1_Int_Ty<1664, "vsubuws" , int_ppc_altivec_vsubuws, v4i32>;
+
+def VSUMSWS : VX1_Int_Ty<1928, "vsumsws" , int_ppc_altivec_vsumsws, v4i32>;
+def VSUM2SWS: VX1_Int_Ty<1672, "vsum2sws", int_ppc_altivec_vsum2sws, v4i32>;
+
+def VSUM4SBS: VX1_Int_Ty3<1672, "vsum4sbs", int_ppc_altivec_vsum4sbs,
+ v4i32, v16i8, v4i32>;
+def VSUM4SHS: VX1_Int_Ty3<1608, "vsum4shs", int_ppc_altivec_vsum4shs,
+ v4i32, v8i16, v4i32>;
+def VSUM4UBS: VX1_Int_Ty3<1544, "vsum4ubs", int_ppc_altivec_vsum4ubs,
+ v4i32, v16i8, v4i32>;
def VNOR : VXForm_1<1284, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vnor $vD, $vA, $vB", VecFP,
@@ -463,15 +550,16 @@ def VXOR : VXForm_1<1220, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vxor $vD, $vA, $vB", VecFP,
[(set VRRC:$vD, (xor (v4i32 VRRC:$vA), VRRC:$vB))]>;
-def VRLB : VX1_Int< 4, "vrlb", int_ppc_altivec_vrlb>;
-def VRLH : VX1_Int< 68, "vrlh", int_ppc_altivec_vrlh>;
-def VRLW : VX1_Int< 132, "vrlw", int_ppc_altivec_vrlw>;
+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< 452, "vsl" , int_ppc_altivec_vsl >;
-def VSLO : VX1_Int<1036, "vslo", int_ppc_altivec_vslo>;
-def VSLB : VX1_Int< 260, "vslb", int_ppc_altivec_vslb>;
-def VSLH : VX1_Int< 324, "vslh", int_ppc_altivec_vslh>;
-def VSLW : VX1_Int< 388, "vslw", int_ppc_altivec_vslw>;
+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, $UIMM", VecPerm,
@@ -486,60 +574,74 @@ def VSPLTW : VXForm_1<652, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB),
[(set VRRC:$vD,
(vspltw_shuffle:$UIMM (v16i8 VRRC:$vB), (undef)))]>;
-def VSR : VX1_Int< 708, "vsr" , int_ppc_altivec_vsr>;
-def VSRO : VX1_Int<1100, "vsro" , int_ppc_altivec_vsro>;
-def VSRAB : VX1_Int< 772, "vsrab", int_ppc_altivec_vsrab>;
-def VSRAH : VX1_Int< 836, "vsrah", int_ppc_altivec_vsrah>;
-def VSRAW : VX1_Int< 900, "vsraw", int_ppc_altivec_vsraw>;
-def VSRB : VX1_Int< 516, "vsrb" , int_ppc_altivec_vsrb>;
-def VSRH : VX1_Int< 580, "vsrh" , int_ppc_altivec_vsrh>;
-def VSRW : VX1_Int< 644, "vsrw" , int_ppc_altivec_vsrw>;
+def VSR : VX1_Int_Ty< 708, "vsr" , int_ppc_altivec_vsr, v4i32>;
+def VSRO : VX1_Int_Ty<1100, "vsro" , int_ppc_altivec_vsro, v4i32>;
+
+def VSRAB : VX1_Int_Ty< 772, "vsrab", int_ppc_altivec_vsrab, v16i8>;
+def VSRAH : VX1_Int_Ty< 836, "vsrah", int_ppc_altivec_vsrah, v8i16>;
+def VSRAW : VX1_Int_Ty< 900, "vsraw", int_ppc_altivec_vsraw, v4i32>;
+def VSRB : VX1_Int_Ty< 516, "vsrb" , int_ppc_altivec_vsrb , v16i8>;
+def VSRH : VX1_Int_Ty< 580, "vsrh" , int_ppc_altivec_vsrh , v8i16>;
+def VSRW : VX1_Int_Ty< 644, "vsrw" , int_ppc_altivec_vsrw , v4i32>;
def VSPLTISB : VXForm_3<780, (outs VRRC:$vD), (ins s5imm:$SIMM),
"vspltisb $vD, $SIMM", VecPerm,
- [(set VRRC:$vD, (v16i8 vecspltisb:$SIMM))]>;
+ [(set v16i8:$vD, (v16i8 vecspltisb:$SIMM))]>;
def VSPLTISH : VXForm_3<844, (outs VRRC:$vD), (ins s5imm:$SIMM),
"vspltish $vD, $SIMM", VecPerm,
- [(set VRRC:$vD, (v8i16 vecspltish:$SIMM))]>;
+ [(set v8i16:$vD, (v8i16 vecspltish:$SIMM))]>;
def VSPLTISW : VXForm_3<908, (outs VRRC:$vD), (ins s5imm:$SIMM),
"vspltisw $vD, $SIMM", VecPerm,
- [(set VRRC:$vD, (v4i32 vecspltisw:$SIMM))]>;
+ [(set v4i32:$vD, (v4i32 vecspltisw:$SIMM))]>;
// Vector Pack.
-def VPKPX : VX1_Int<782, "vpkpx", int_ppc_altivec_vpkpx>;
-def VPKSHSS : VX1_Int<398, "vpkshss", int_ppc_altivec_vpkshss>;
-def VPKSHUS : VX1_Int<270, "vpkshus", int_ppc_altivec_vpkshus>;
-def VPKSWSS : VX1_Int<462, "vpkswss", int_ppc_altivec_vpkswss>;
-def VPKSWUS : VX1_Int<334, "vpkswus", int_ppc_altivec_vpkswus>;
+def VPKPX : VX1_Int_Ty2<782, "vpkpx", int_ppc_altivec_vpkpx,
+ v8i16, v4i32>;
+def VPKSHSS : VX1_Int_Ty2<398, "vpkshss", int_ppc_altivec_vpkshss,
+ v16i8, v8i16>;
+def VPKSHUS : VX1_Int_Ty2<270, "vpkshus", int_ppc_altivec_vpkshus,
+ v16i8, v8i16>;
+def VPKSWSS : VX1_Int_Ty2<462, "vpkswss", int_ppc_altivec_vpkswss,
+ v16i8, v4i32>;
+def VPKSWUS : VX1_Int_Ty2<334, "vpkswus", int_ppc_altivec_vpkswus,
+ v8i16, v4i32>;
def VPKUHUM : VXForm_1<14, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),
"vpkuhum $vD, $vA, $vB", VecFP,
[(set VRRC:$vD,
(vpkuhum_shuffle (v16i8 VRRC:$vA), VRRC:$vB))]>;
-def VPKUHUS : VX1_Int<142, "vpkuhus", int_ppc_altivec_vpkuhus>;
+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, $vB", VecFP,
[(set VRRC:$vD,
(vpkuwum_shuffle (v16i8 VRRC:$vA), VRRC:$vB))]>;
-def VPKUWUS : VX1_Int<206, "vpkuwus", int_ppc_altivec_vpkuwus>;
+def VPKUWUS : VX1_Int_Ty2<206, "vpkuwus", int_ppc_altivec_vpkuwus,
+ v8i16, v4i32>;
// Vector Unpack.
-def VUPKHPX : VX2_Int<846, "vupkhpx", int_ppc_altivec_vupkhpx>;
-def VUPKHSB : VX2_Int<526, "vupkhsb", int_ppc_altivec_vupkhsb>;
-def VUPKHSH : VX2_Int<590, "vupkhsh", int_ppc_altivec_vupkhsh>;
-def VUPKLPX : VX2_Int<974, "vupklpx", int_ppc_altivec_vupklpx>;
-def VUPKLSB : VX2_Int<654, "vupklsb", int_ppc_altivec_vupklsb>;
-def VUPKLSH : VX2_Int<718, "vupklsh", int_ppc_altivec_vupklsh>;
+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<bits<10> xo, string asmstr, ValueType Ty>
: VXRForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),asmstr,VecFPCompare,
- [(set VRRC:$vD, (Ty (PPCvcmp VRRC:$vA, VRRC:$vB, xo)))]>;
+ [(set Ty:$vD, (Ty (PPCvcmp Ty:$vA, Ty:$vB, xo)))]>;
class VCMPo<bits<10> xo, string asmstr, ValueType Ty>
: VXRForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),asmstr,VecFPCompare,
- [(set VRRC:$vD, (Ty (PPCvcmp_o VRRC:$vA, VRRC:$vB, xo)))]> {
+ [(set Ty:$vD, (Ty (PPCvcmp_o Ty:$vA, Ty:$vB, xo)))]> {
let Defs = [CR6];
let RC = 1;
}
@@ -578,13 +680,14 @@ 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, $vB", v4i32>;
+let isCodeGenOnly = 1 in
def V_SET0 : VXForm_setzero<1220, (outs VRRC:$vD), (ins),
"vxor $vD, $vD, $vD", VecFP,
[(set VRRC:$vD, (v4i32 immAllZerosV))]>;
let IMM=-1 in {
def V_SETALLONES : VXForm_3<908, (outs VRRC:$vD), (ins),
"vspltisw $vD, -1", VecFP,
- [(set VRRC:$vD, (v4i32 immAllOnesV))]>;
+ [(set v4i32:$vD, (v4i32 immAllOnesV))]>;
}
} // VALU Operations.
@@ -597,31 +700,31 @@ def : Pat<(int_ppc_altivec_dssall), (DSSALL 1, 0, 0, 0)>;
def : Pat<(int_ppc_altivec_dss imm:$STRM), (DSS 0, imm:$STRM, 0, 0)>;
// * 32-bit
-def : Pat<(int_ppc_altivec_dst GPRC:$rA, GPRC:$rB, imm:$STRM),
- (DST 0, imm:$STRM, GPRC:$rA, GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dstt GPRC:$rA, GPRC:$rB, imm:$STRM),
- (DSTT 1, imm:$STRM, GPRC:$rA, GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dstst GPRC:$rA, GPRC:$rB, imm:$STRM),
- (DSTST 0, imm:$STRM, GPRC:$rA, GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dststt GPRC:$rA, GPRC:$rB, imm:$STRM),
- (DSTSTT 1, imm:$STRM, GPRC:$rA, GPRC:$rB)>;
+def : Pat<(int_ppc_altivec_dst i32:$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, $rB)>;
// * 64-bit
-def : Pat<(int_ppc_altivec_dst G8RC:$rA, GPRC:$rB, imm:$STRM),
- (DST64 0, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dstt G8RC:$rA, GPRC:$rB, imm:$STRM),
- (DSTT64 1, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dstst G8RC:$rA, GPRC:$rB, imm:$STRM),
- (DSTST64 0, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>;
-def : Pat<(int_ppc_altivec_dststt G8RC:$rA, GPRC:$rB, imm:$STRM),
- (DSTSTT64 1, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>;
+def : Pat<(int_ppc_altivec_dst i64:$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:$src)>;
// Stores.
-def : Pat<(store (v4i32 VRRC:$rS), xoaddr:$dst),
- (STVX (v4i32 VRRC:$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)>;
@@ -643,93 +746,96 @@ def : Pat<(v4f32 (bitconvert (v4i32 VRRC:$src))), (v4f32 VRRC:$src)>;
// Shuffles.
// Match vsldoi(x,x), vpkuwum(x,x), vpkuhum(x,x)
-def:Pat<(vsldoi_unary_shuffle:$in (v16i8 VRRC:$vA), undef),
- (VSLDOI VRRC:$vA, VRRC:$vA, (VSLDOI_unary_get_imm VRRC:$in))>;
-def:Pat<(vpkuwum_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VPKUWUM VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vpkuhum_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VPKUHUM VRRC:$vA, VRRC:$vA)>;
+def:Pat<(vsldoi_unary_shuffle:$in v16i8:$vA, undef),
+ (VSLDOI $vA, $vA, (VSLDOI_unary_get_imm VRRC:$in))>;
+def:Pat<(vpkuwum_unary_shuffle v16i8:$vA, undef),
+ (VPKUWUM $vA, $vA)>;
+def:Pat<(vpkuhum_unary_shuffle v16i8:$vA, undef),
+ (VPKUHUM $vA, $vA)>;
// Match vmrg*(x,x)
-def:Pat<(vmrglb_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGLB VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vmrglh_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGLH VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vmrglw_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGLW VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vmrghb_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGHB VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vmrghh_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGHH VRRC:$vA, VRRC:$vA)>;
-def:Pat<(vmrghw_unary_shuffle (v16i8 VRRC:$vA), undef),
- (VMRGHW VRRC:$vA, VRRC:$vA)>;
+def:Pat<(vmrglb_unary_shuffle v16i8:$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, $vA)>;
// Logical Operations
-def : Pat<(v4i32 (vnot_ppc VRRC:$vA)), (VNOR VRRC:$vA, VRRC:$vA)>;
+def : Pat<(v4i32 (vnot_ppc VRRC:$vA)), (VNOR $vA, $vA)>;
def : Pat<(v4i32 (vnot_ppc (or VRRC:$A, VRRC:$B))),
- (VNOR VRRC:$A, VRRC:$B)>;
+ (VNOR $A, $B)>;
def : Pat<(v4i32 (and VRRC:$A, (vnot_ppc VRRC:$B))),
- (VANDC VRRC:$A, VRRC:$B)>;
+ (VANDC $A, $B)>;
-def : Pat<(fmul VRRC:$vA, VRRC:$vB),
- (VMADDFP VRRC:$vA, VRRC:$vB,
+def : Pat<(fmul v4f32:$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:$C)>;
+def : Pat<(PPCvmaddfp v4f32:$A, v4f32:$B, v4f32:$C),
+ (VMADDFP $A, $B, $C)>;
+def : Pat<(PPCvnmsubfp v4f32:$A, v4f32:$B, v4f32:$C),
+ (VNMSUBFP $A, $B, $C)>;
-def : Pat<(int_ppc_altivec_vmaddfp VRRC:$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:$C)>;
+def : Pat<(int_ppc_altivec_vmaddfp v4f32:$A, v4f32:$B, v4f32:$C),
+ (VMADDFP $A, $B, $C)>;
+def : Pat<(int_ppc_altivec_vnmsubfp v4f32:$A, v4f32:$B, v4f32:$C),
+ (VNMSUBFP $A, $B, $C)>;
def : Pat<(PPCvperm (v16i8 VRRC:$vA), VRRC:$vB, VRRC:$vC),
- (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC)>;
+ (VPERM $vA, $vB, $vC)>;
// Vector shifts
-def : Pat<(v16i8 (shl (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))),
- (v16i8 (VSLB VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v8i16 (shl (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))),
- (v8i16 (VSLH VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v4i32 (shl (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))),
- (v4i32 (VSLW VRRC:$vA, VRRC:$vB))>;
-
-def : Pat<(v16i8 (srl (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))),
- (v16i8 (VSRB VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v8i16 (srl (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))),
- (v8i16 (VSRH VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v4i32 (srl (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))),
- (v4i32 (VSRW VRRC:$vA, VRRC:$vB))>;
-
-def : Pat<(v16i8 (sra (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))),
- (v16i8 (VSRAB VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v8i16 (sra (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))),
- (v8i16 (VSRAH VRRC:$vA, VRRC:$vB))>;
-def : Pat<(v4i32 (sra (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))),
- (v4i32 (VSRAW VRRC:$vA, VRRC:$vB))>;
+def : Pat<(v16i8 (shl v16i8:$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, $vB))>;
// Float to integer and integer to float conversions
-def : Pat<(v4i32 (fp_to_sint (v4f32 VRRC:$vA))),
- (VCTSXS_0 VRRC:$vA)>;
-def : Pat<(v4i32 (fp_to_uint (v4f32 VRRC:$vA))),
- (VCTUXS_0 VRRC:$vA)>;
-def : Pat<(v4f32 (sint_to_fp (v4i32 VRRC:$vA))),
- (VCFSX_0 VRRC:$vA)>;
-def : Pat<(v4f32 (uint_to_fp (v4i32 VRRC:$vA))),
- (VCFUX_0 VRRC:$vA)>;
+def : Pat<(v4i32 (fp_to_sint v4f32:$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 $vA)>;
// Floating-point rounding
-def : Pat<(v4f32 (ffloor (v4f32 VRRC:$vA))),
- (VRFIM VRRC:$vA)>;
-def : Pat<(v4f32 (fceil (v4f32 VRRC:$vA))),
- (VRFIP VRRC:$vA)>;
-def : Pat<(v4f32 (ftrunc (v4f32 VRRC:$vA))),
- (VRFIZ VRRC:$vA)>;
-def : Pat<(v4f32 (fnearbyint (v4f32 VRRC:$vA))),
- (VRFIN VRRC:$vA)>;
+def : Pat<(v4f32 (ffloor v4f32:$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
+