[llvm] [PowerPC] Prefer VSX compare in vector selection (PR #65295)
Qiu Chaofan via llvm-commits
llvm-commits at lists.llvm.org
Sun Dec 3 23:15:12 PST 2023
https://github.com/ecnelises updated https://github.com/llvm/llvm-project/pull/65295
>From 6f576aa2fc16856c4c920e23a17f68fa687845c3 Mon Sep 17 00:00:00 2001
From: Qiu Chaofan <qiucofan at cn.ibm.com>
Date: Mon, 4 Dec 2023 15:14:23 +0800
Subject: [PATCH] Use VSX compare and xxsel to optimize float select_cc
---
llvm/lib/Target/PowerPC/PPCISelLowering.cpp | 63 ++++
llvm/lib/Target/PowerPC/PPCISelLowering.h | 3 +
llvm/lib/Target/PowerPC/PPCInstrInfo.td | 2 +
llvm/lib/Target/PowerPC/PPCInstrP10.td | 13 +
llvm/lib/Target/PowerPC/PPCInstrVSX.td | 46 +++
llvm/test/CodeGen/PowerPC/cgp-select.ll | 30 +-
llvm/test/CodeGen/PowerPC/fmf-propagation.ll | 224 ++++++------
llvm/test/CodeGen/PowerPC/recipest.ll | 52 ++-
llvm/test/CodeGen/PowerPC/scalar-equal.ll | 24 +-
llvm/test/CodeGen/PowerPC/scalar_cmp.ll | 358 +++++++++----------
llvm/test/CodeGen/PowerPC/vsx.ll | 33 +-
11 files changed, 484 insertions(+), 364 deletions(-)
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index ca65048e2355f..4365f6fccaaa9 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -1830,6 +1830,8 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const {
case PPCISD::LXVRZX: return "PPCISD::LXVRZX";
case PPCISD::STORE_COND:
return "PPCISD::STORE_COND";
+ case PPCISD::VSX_CMPSEL:
+ return "PPCISD::VSX_CMPSEL";
}
return nullptr;
}
@@ -15560,6 +15562,65 @@ static bool isStoreConditional(SDValue Intrin, unsigned &StoreWidth) {
return true;
}
+// Use VSX compare gt/ge/eq instruction to implement select_cc
+static SDValue combineFloatSelectCC(SDValue Op, const PPCSubtarget &Subtarget,
+ SelectionDAG &DAG) {
+ EVT VT = Op.getValueType();
+
+ // Use subtraction based lowering if it is finite-math.
+ if (DAG.getTarget().Options.NoInfsFPMath || Op->getFlags().hasNoInfs())
+ return SDValue();
+
+ // Vector comparison is already implemented in isel.
+ if (VT.isVector() || !VT.isFloatingPoint() ||
+ (VT == MVT::f128 && !Subtarget.hasP10Vector()) || !Subtarget.hasVSX())
+ return SDValue();
+ SDValue TrueVal = Op.getOperand(2), FalseVal = Op.getOperand(3);
+ SDValue Cond1 = Op.getOperand(0), Cond2 = Op.getOperand(1);
+ SDLoc DL(Op);
+ unsigned CC = cast<CondCodeSDNode>(Op.getOperand(4))->get();
+
+ // Recognize the form after legalizer if cond code is illegal.
+ if (VT != Cond1.getValueType()) {
+ if (!isNullConstant(Cond2) || CC != ISD::SETNE ||
+ Cond1.getOpcode() != ISD::AND)
+ return SDValue();
+ Cond1 = Cond1.getOperand(0);
+ Cond2 = Cond1.getOperand(1);
+ if (Cond1.getOpcode() != ISD::SETCC || Cond2.getOpcode() != ISD::SETCC ||
+ Cond1.getOperand(0) != Cond2.getOperand(0) ||
+ Cond1.getOperand(1) != Cond2.getOperand(1))
+ return SDValue();
+ CC = cast<CondCodeSDNode>(Cond1.getOperand(2))->get() &
+ cast<CondCodeSDNode>(Cond2.getOperand(2))->get();
+ Cond1 = Cond1.getOperand(0);
+ Cond2 = Cond1.getOperand(1);
+ }
+
+ // The instruction is ordered. Treat it as ordered if we don't care order.
+ if (CC & ISD::SETUO)
+ return SDValue();
+ if (CC & ISD::SETFALSE2)
+ CC &= ISD::SETO;
+
+ // Use min/max instructions if available.
+ if (((Cond1 == TrueVal && Cond2 == FalseVal) ||
+ (Cond1 == FalseVal && Cond2 == TrueVal)) &&
+ (CC == ISD::SETOLT || CC == ISD::SETOGT))
+ return SDValue();
+
+ bool Inverse = false;
+ if (CC == ISD::SETOLT || CC == ISD::SETOLE || CC == ISD::SETONE) {
+ CC = (~CC) & ISD::SETO;
+ Inverse = true;
+ }
+ if (CC == ISD::SETOGE || CC == ISD::SETOGT || CC == ISD::SETOEQ)
+ return DAG.getNode(PPCISD::VSX_CMPSEL, DL, VT, Inverse ? Cond2 : Cond1,
+ Inverse ? Cond1 : Cond2, TrueVal, FalseVal,
+ DAG.getCondCode((ISD::CondCode)CC));
+ return SDValue();
+}
+
SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N,
DAGCombinerInfo &DCI) const {
SelectionDAG &DAG = DCI.DAG;
@@ -15629,6 +15690,8 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N,
return CSCC;
[[fallthrough]];
case ISD::SELECT_CC:
+ if (SDValue V = combineFloatSelectCC(SDValue(N, 0), Subtarget, DCI.DAG))
+ return V;
return DAGCombineTruncBoolExt(N, DCI);
case ISD::SINT_TO_FP:
case ISD::UINT_TO_FP:
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.h b/llvm/lib/Target/PowerPC/PPCISelLowering.h
index d8679dcf40180..100a73f99f493 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.h
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.h
@@ -475,6 +475,9 @@ namespace llvm {
/// XXMFACC = This corresponds to the xxmfacc instruction.
XXMFACC,
+ /// VSX_CMPSEL = VSX compare gt/ge/eq instruction with selection.
+ VSX_CMPSEL,
+
// Constrained conversion from floating point to int
STRICT_FCTIDZ = ISD::FIRST_TARGET_STRICTFP_OPCODE,
STRICT_FCTIWZ,
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.td b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
index a97062e0c643f..c527f71e1cec7 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
@@ -193,6 +193,8 @@ def PPCany_faddrtz: PatFrags<(ops node:$lhs, node:$rhs),
[(PPCfaddrtz node:$lhs, node:$rhs),
(PPCstrict_faddrtz node:$lhs, node:$rhs)]>;
+def PPCvsx_cmpsel : SDNode<"PPCISD::VSX_CMPSEL", SDTSelectCC, []>;
+
def PPCfsel : SDNode<"PPCISD::FSEL",
// Type constraint for fsel.
SDTypeProfile<1, 3, [SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>,
diff --git a/llvm/lib/Target/PowerPC/PPCInstrP10.td b/llvm/lib/Target/PowerPC/PPCInstrP10.td
index f72a52d5b26f6..df188da4fa9d3 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrP10.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrP10.td
@@ -2067,6 +2067,19 @@ let AddedComplexity = 400, Predicates = [IsISA3_1, HasVSX] in {
(v1i128 (VSRAQ v1i128:$VRA,
(XXPERMDI (COPY_TO_REGCLASS $VRB, VSRC),
(COPY_TO_REGCLASS $VRB, VSRC), 2)))>;
+
+def : Pat<(f128 (PPCvsx_cmpsel f128:$lhs, f128:$rhs, f128:$tval, f128:$fval, SETOEQ)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPEQQP $lhs, $rhs)), VSFRC)>;
+def : Pat<(f128 (PPCvsx_cmpsel f128:$lhs, f128:$rhs, f128:$tval, f128:$fval, SETOGE)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPGEQP $lhs, $rhs)), VSFRC)>;
+def : Pat<(f128 (PPCvsx_cmpsel f128:$lhs, f128:$rhs, f128:$tval, f128:$fval, SETOGT)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPGTQP $lhs, $rhs)), VSFRC)>;
}
class xxevalPattern <dag pattern, bits<8> imm> :
diff --git a/llvm/lib/Target/PowerPC/PPCInstrVSX.td b/llvm/lib/Target/PowerPC/PPCInstrVSX.td
index 0e5f6b773bb54..58a362f5d8af4 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrVSX.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrVSX.td
@@ -2918,6 +2918,39 @@ def : Pat<(PPCstore_scal_int_from_vsr f64:$src, XForm:$dst, 8),
(STXSDX $src, XForm:$dst)>;
def : Pat<(PPCstore_scal_int_from_vsr f128:$src, XForm:$dst, 8),
(STXSDX (COPY_TO_REGCLASS $src, VSFRC), XForm:$dst)>;
+
+def : Pat<(f32 (PPCvsx_cmpsel f32:$lhs, f32:$rhs, f32:$tval, f32:$fval, SETOEQ)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC), (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPEQSP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSSRC)>;
+def : Pat<(f32 (PPCvsx_cmpsel f32:$lhs, f32:$rhs, f32:$tval, f32:$fval, SETOGE)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC), (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPGESP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSSRC)>;
+def : Pat<(f32 (PPCvsx_cmpsel f32:$lhs, f32:$rhs, f32:$tval, f32:$fval, SETOGT)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPGTSP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSSRC)>;
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOEQ)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC), (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPEQDP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSFRC)>;
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOGE)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC), (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPGEDP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSFRC)>;
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOGT)),
+ (COPY_TO_REGCLASS (XXSEL
+ (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XVCMPGTDP (COPY_TO_REGCLASS $lhs, VSRC),
+ (COPY_TO_REGCLASS $rhs, VSRC))), VSFRC)>;
} // HasVSX
// Any big endian VSX subtarget.
@@ -3933,6 +3966,19 @@ foreach Ty = [v4i32, v4f32, v2i64, v2f64] in {
def : Pat<(store Ty:$rS, XForm:$dst), (STXVX $rS, XForm:$dst)>;
}
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOEQ)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPEQDP $lhs, $rhs)), VSFRC)>;
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOGE)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPGEDP $lhs, $rhs)), VSFRC)>;
+def : Pat<(f64 (PPCvsx_cmpsel f64:$lhs, f64:$rhs, f64:$tval, f64:$fval, SETOGT)),
+ (COPY_TO_REGCLASS (XXSEL (COPY_TO_REGCLASS $tval, VSRC),
+ (COPY_TO_REGCLASS $fval, VSRC),
+ (XSCMPGTDP $lhs, $rhs)), VSFRC)>;
+
def : Pat<(f128 (load DQForm:$src)),
(COPY_TO_REGCLASS (LXV memrix16:$src), VRRC)>;
def : Pat<(f128 (load XForm:$src)),
diff --git a/llvm/test/CodeGen/PowerPC/cgp-select.ll b/llvm/test/CodeGen/PowerPC/cgp-select.ll
index 2c36233b2b714..dc57c505c6bc4 100644
--- a/llvm/test/CodeGen/PowerPC/cgp-select.ll
+++ b/llvm/test/CodeGen/PowerPC/cgp-select.ll
@@ -8,30 +8,32 @@ define dso_local void @wibble(ptr nocapture readonly %arg, i32 signext %arg1, pt
; CHECK-NEXT: li 7, 7
; CHECK-NEXT: cmpwi 4, 2
; CHECK-NEXT: xsaddsp 0, 0, 0
-; CHECK-NEXT: blt 0, .LBB0_5
+; CHECK-NEXT: blt 0, .LBB0_4
; CHECK-NEXT: # %bb.1: # %bb6
; CHECK-NEXT: clrldi 4, 4, 32
; CHECK-NEXT: addi 4, 4, -1
; CHECK-NEXT: mtctr 4
; CHECK-NEXT: li 4, 8
-; CHECK-NEXT: b .LBB0_3
-; CHECK-NEXT: .p2align 5
+; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB0_2: # %bb11
; CHECK-NEXT: #
+; CHECK-NEXT: lfsu 2, 4(3)
+; CHECK-NEXT: xsaddsp 1, 2, 2
+; CHECK-NEXT: xvcmpgtsp 3, 2, 0
+; CHECK-NEXT: fcmpu 0, 2, 0
; CHECK-NEXT: iselgt 7, 4, 7
; CHECK-NEXT: addi 4, 4, 1
-; CHECK-NEXT: bdz .LBB0_5
-; CHECK-NEXT: .LBB0_3: # %bb11
-; CHECK-NEXT: #
-; CHECK-NEXT: lfsu 1, 4(3)
-; CHECK-NEXT: fcmpu 0, 1, 0
-; CHECK-NEXT: ble 0, .LBB0_2
-; CHECK-NEXT: # %bb.4:
-; CHECK-NEXT: xsaddsp 0, 1, 1
-; CHECK-NEXT: b .LBB0_2
-; CHECK-NEXT: .LBB0_5: # %bb8
+; CHECK-NEXT: xxsel 1, 1, 0, 3
+; CHECK-NEXT: fmr 0, 1
+; CHECK-NEXT: bdnz .LBB0_2
+; CHECK-NEXT: # %bb.3: # %bb8
+; CHECK-NEXT: stw 7, 0(5)
+; CHECK-NEXT: stfs 1, 0(6)
+; CHECK-NEXT: blr
+; CHECK-NEXT: .LBB0_4:
+; CHECK-NEXT: fmr 1, 0
; CHECK-NEXT: stw 7, 0(5)
-; CHECK-NEXT: stfs 0, 0(6)
+; CHECK-NEXT: stfs 1, 0(6)
; CHECK-NEXT: blr
bb:
%tmp = load float, ptr %arg, align 4
diff --git a/llvm/test/CodeGen/PowerPC/fmf-propagation.ll b/llvm/test/CodeGen/PowerPC/fmf-propagation.ll
index 58b3ee485ea4b..16f2df13e78a4 100644
--- a/llvm/test/CodeGen/PowerPC/fmf-propagation.ll
+++ b/llvm/test/CodeGen/PowerPC/fmf-propagation.ll
@@ -302,47 +302,43 @@ define float @fmul_fma_fast2(float %x) {
define float @sqrt_afn_ieee(float %x) #0 {
; FMF-LABEL: sqrt_afn_ieee:
; FMF: # %bb.0:
-; FMF-NEXT: addis 3, 2, .LCPI11_1 at toc@ha
-; FMF-NEXT: xsabsdp 0, 1
-; FMF-NEXT: lfs 2, .LCPI11_1 at toc@l(3)
-; FMF-NEXT: fcmpu 0, 0, 2
-; FMF-NEXT: xxlxor 0, 0, 0
-; FMF-NEXT: blt 0, .LBB11_2
-; FMF-NEXT: # %bb.1:
; FMF-NEXT: xsrsqrtesp 2, 1
; FMF-NEXT: addis 3, 2, .LCPI11_0 at toc@ha
; FMF-NEXT: vspltisw 2, -3
; FMF-NEXT: lfs 0, .LCPI11_0 at toc@l(3)
-; FMF-NEXT: xsmulsp 1, 1, 2
-; FMF-NEXT: xsmulsp 0, 1, 0
-; FMF-NEXT: xsmulsp 1, 1, 2
-; FMF-NEXT: xvcvsxwdp 2, 34
-; FMF-NEXT: xsaddsp 1, 1, 2
-; FMF-NEXT: xsmulsp 0, 0, 1
-; FMF-NEXT: .LBB11_2:
-; FMF-NEXT: fmr 1, 0
+; FMF-NEXT: addis 3, 2, .LCPI11_1 at toc@ha
+; FMF-NEXT: xsmulsp 3, 1, 2
+; FMF-NEXT: xsabsdp 1, 1
+; FMF-NEXT: xsmulsp 0, 3, 0
+; FMF-NEXT: xsmulsp 2, 3, 2
+; FMF-NEXT: xvcvsxwdp 3, 34
+; FMF-NEXT: xsaddsp 2, 2, 3
+; FMF-NEXT: xsmulsp 0, 0, 2
+; FMF-NEXT: lfs 2, .LCPI11_1 at toc@l(3)
+; FMF-NEXT: xvcmpgesp 1, 2, 1
+; FMF-NEXT: xxlxor 2, 2, 2
+; FMF-NEXT: xxsel 1, 2, 0, 1
+; FMF-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; FMF-NEXT: blr
;
; GLOBAL-LABEL: sqrt_afn_ieee:
; GLOBAL: # %bb.0:
-; GLOBAL-NEXT: addis 3, 2, .LCPI11_1 at toc@ha
-; GLOBAL-NEXT: xsabsdp 0, 1
-; GLOBAL-NEXT: lfs 2, .LCPI11_1 at toc@l(3)
-; GLOBAL-NEXT: fcmpu 0, 0, 2
-; GLOBAL-NEXT: xxlxor 0, 0, 0
-; GLOBAL-NEXT: blt 0, .LBB11_2
-; GLOBAL-NEXT: # %bb.1:
; GLOBAL-NEXT: xsrsqrtesp 0, 1
; GLOBAL-NEXT: vspltisw 2, -3
; GLOBAL-NEXT: addis 3, 2, .LCPI11_0 at toc@ha
-; GLOBAL-NEXT: xvcvsxwdp 2, 34
-; GLOBAL-NEXT: xsmulsp 1, 1, 0
-; GLOBAL-NEXT: xsmaddasp 2, 1, 0
+; GLOBAL-NEXT: xvcvsxwdp 3, 34
+; GLOBAL-NEXT: xsmulsp 2, 1, 0
+; GLOBAL-NEXT: xsabsdp 1, 1
+; GLOBAL-NEXT: xsmaddasp 3, 2, 0
; GLOBAL-NEXT: lfs 0, .LCPI11_0 at toc@l(3)
-; GLOBAL-NEXT: xsmulsp 0, 1, 0
-; GLOBAL-NEXT: xsmulsp 0, 0, 2
-; GLOBAL-NEXT: .LBB11_2:
-; GLOBAL-NEXT: fmr 1, 0
+; GLOBAL-NEXT: addis 3, 2, .LCPI11_1 at toc@ha
+; GLOBAL-NEXT: xsmulsp 0, 2, 0
+; GLOBAL-NEXT: lfs 2, .LCPI11_1 at toc@l(3)
+; GLOBAL-NEXT: xvcmpgesp 1, 2, 1
+; GLOBAL-NEXT: xxlxor 2, 2, 2
+; GLOBAL-NEXT: xsmulsp 0, 0, 3
+; GLOBAL-NEXT: xxsel 1, 2, 0, 1
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; GLOBAL-NEXT: blr
%rt = call afn ninf float @llvm.sqrt.f32(float %x)
ret float %rt
@@ -373,41 +369,39 @@ define float @sqrt_afn_ieee_inf(float %x) #0 {
define float @sqrt_afn_preserve_sign(float %x) #1 {
; FMF-LABEL: sqrt_afn_preserve_sign:
; FMF: # %bb.0:
-; FMF-NEXT: xxlxor 0, 0, 0
-; FMF-NEXT: fcmpu 0, 1, 0
-; FMF-NEXT: beq 0, .LBB13_2
-; FMF-NEXT: # %bb.1:
-; FMF-NEXT: xsrsqrtesp 0, 1
+; FMF-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; FMF-NEXT: xsrsqrtesp 2, 1
; FMF-NEXT: addis 3, 2, .LCPI13_0 at toc@ha
; FMF-NEXT: vspltisw 2, -3
-; FMF-NEXT: lfs 2, .LCPI13_0 at toc@l(3)
-; FMF-NEXT: xsmulsp 1, 1, 0
-; FMF-NEXT: xsmulsp 2, 1, 2
-; FMF-NEXT: xsmulsp 0, 1, 0
-; FMF-NEXT: xvcvsxwdp 1, 34
-; FMF-NEXT: xsaddsp 0, 0, 1
-; FMF-NEXT: xsmulsp 0, 2, 0
-; FMF-NEXT: .LBB13_2:
-; FMF-NEXT: fmr 1, 0
+; FMF-NEXT: lfs 0, .LCPI13_0 at toc@l(3)
+; FMF-NEXT: xsmulsp 3, 1, 2
+; FMF-NEXT: xsmulsp 0, 3, 0
+; FMF-NEXT: xsmulsp 2, 3, 2
+; FMF-NEXT: xvcvsxwdp 3, 34
+; FMF-NEXT: xsaddsp 2, 2, 3
+; FMF-NEXT: xsmulsp 0, 0, 2
+; FMF-NEXT: xxlxor 2, 2, 2
+; FMF-NEXT: xvcmpeqsp 1, 1, 2
+; FMF-NEXT: xxsel 1, 2, 0, 1
+; FMF-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; FMF-NEXT: blr
;
; GLOBAL-LABEL: sqrt_afn_preserve_sign:
; GLOBAL: # %bb.0:
-; GLOBAL-NEXT: xxlxor 0, 0, 0
-; GLOBAL-NEXT: fcmpu 0, 1, 0
-; GLOBAL-NEXT: beq 0, .LBB13_2
-; GLOBAL-NEXT: # %bb.1:
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 def $vsl1
; GLOBAL-NEXT: xsrsqrtesp 0, 1
; GLOBAL-NEXT: vspltisw 2, -3
; GLOBAL-NEXT: addis 3, 2, .LCPI13_0 at toc@ha
-; GLOBAL-NEXT: xvcvsxwdp 2, 34
-; GLOBAL-NEXT: xsmulsp 1, 1, 0
-; GLOBAL-NEXT: xsmaddasp 2, 1, 0
+; GLOBAL-NEXT: xvcvsxwdp 3, 34
+; GLOBAL-NEXT: xsmulsp 2, 1, 0
+; GLOBAL-NEXT: xsmaddasp 3, 2, 0
; GLOBAL-NEXT: lfs 0, .LCPI13_0 at toc@l(3)
-; GLOBAL-NEXT: xsmulsp 0, 1, 0
-; GLOBAL-NEXT: xsmulsp 0, 0, 2
-; GLOBAL-NEXT: .LBB13_2:
-; GLOBAL-NEXT: fmr 1, 0
+; GLOBAL-NEXT: xsmulsp 0, 2, 0
+; GLOBAL-NEXT: xxlxor 2, 2, 2
+; GLOBAL-NEXT: xvcmpeqsp 1, 1, 2
+; GLOBAL-NEXT: xsmulsp 0, 0, 3
+; GLOBAL-NEXT: xxsel 1, 2, 0, 1
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; GLOBAL-NEXT: blr
%rt = call afn ninf float @llvm.sqrt.f32(float %x)
ret float %rt
@@ -440,46 +434,42 @@ define float @sqrt_afn_preserve_sign_inf(float %x) #1 {
define float @sqrt_fast_ieee(float %x) #0 {
; FMF-LABEL: sqrt_fast_ieee:
; FMF: # %bb.0:
-; FMF-NEXT: addis 3, 2, .LCPI15_1 at toc@ha
-; FMF-NEXT: xsabsdp 0, 1
-; FMF-NEXT: lfs 2, .LCPI15_1 at toc@l(3)
-; FMF-NEXT: fcmpu 0, 0, 2
-; FMF-NEXT: xxlxor 0, 0, 0
-; FMF-NEXT: blt 0, .LBB15_2
-; FMF-NEXT: # %bb.1:
; FMF-NEXT: xsrsqrtesp 0, 1
; FMF-NEXT: vspltisw 2, -3
; FMF-NEXT: addis 3, 2, .LCPI15_0 at toc@ha
-; FMF-NEXT: xvcvsxwdp 2, 34
-; FMF-NEXT: xsmulsp 1, 1, 0
-; FMF-NEXT: xsmaddasp 2, 1, 0
+; FMF-NEXT: xvcvsxwdp 3, 34
+; FMF-NEXT: xsmulsp 2, 1, 0
+; FMF-NEXT: xsabsdp 1, 1
+; FMF-NEXT: xsmaddasp 3, 2, 0
; FMF-NEXT: lfs 0, .LCPI15_0 at toc@l(3)
-; FMF-NEXT: xsmulsp 0, 1, 0
-; FMF-NEXT: xsmulsp 0, 0, 2
-; FMF-NEXT: .LBB15_2:
-; FMF-NEXT: fmr 1, 0
+; FMF-NEXT: addis 3, 2, .LCPI15_1 at toc@ha
+; FMF-NEXT: xsmulsp 0, 2, 0
+; FMF-NEXT: lfs 2, .LCPI15_1 at toc@l(3)
+; FMF-NEXT: xvcmpgesp 1, 2, 1
+; FMF-NEXT: xxlxor 2, 2, 2
+; FMF-NEXT: xsmulsp 0, 0, 3
+; FMF-NEXT: xxsel 1, 2, 0, 1
+; FMF-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; FMF-NEXT: blr
;
; GLOBAL-LABEL: sqrt_fast_ieee:
; GLOBAL: # %bb.0:
-; GLOBAL-NEXT: addis 3, 2, .LCPI15_1 at toc@ha
-; GLOBAL-NEXT: xsabsdp 0, 1
-; GLOBAL-NEXT: lfs 2, .LCPI15_1 at toc@l(3)
-; GLOBAL-NEXT: fcmpu 0, 0, 2
-; GLOBAL-NEXT: xxlxor 0, 0, 0
-; GLOBAL-NEXT: blt 0, .LBB15_2
-; GLOBAL-NEXT: # %bb.1:
; GLOBAL-NEXT: xsrsqrtesp 0, 1
; GLOBAL-NEXT: vspltisw 2, -3
; GLOBAL-NEXT: addis 3, 2, .LCPI15_0 at toc@ha
-; GLOBAL-NEXT: xvcvsxwdp 2, 34
-; GLOBAL-NEXT: xsmulsp 1, 1, 0
-; GLOBAL-NEXT: xsmaddasp 2, 1, 0
+; GLOBAL-NEXT: xvcvsxwdp 3, 34
+; GLOBAL-NEXT: xsmulsp 2, 1, 0
+; GLOBAL-NEXT: xsabsdp 1, 1
+; GLOBAL-NEXT: xsmaddasp 3, 2, 0
; GLOBAL-NEXT: lfs 0, .LCPI15_0 at toc@l(3)
-; GLOBAL-NEXT: xsmulsp 0, 1, 0
-; GLOBAL-NEXT: xsmulsp 0, 0, 2
-; GLOBAL-NEXT: .LBB15_2:
-; GLOBAL-NEXT: fmr 1, 0
+; GLOBAL-NEXT: addis 3, 2, .LCPI15_1 at toc@ha
+; GLOBAL-NEXT: xsmulsp 0, 2, 0
+; GLOBAL-NEXT: lfs 2, .LCPI15_1 at toc@l(3)
+; GLOBAL-NEXT: xvcmpgesp 1, 2, 1
+; GLOBAL-NEXT: xxlxor 2, 2, 2
+; GLOBAL-NEXT: xsmulsp 0, 0, 3
+; GLOBAL-NEXT: xxsel 1, 2, 0, 1
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; GLOBAL-NEXT: blr
%rt = call contract reassoc afn ninf float @llvm.sqrt.f32(float %x)
ret float %rt
@@ -498,40 +488,38 @@ define float @sqrt_fast_ieee(float %x) #0 {
define float @sqrt_fast_preserve_sign(float %x) #1 {
; FMF-LABEL: sqrt_fast_preserve_sign:
; FMF: # %bb.0:
-; FMF-NEXT: xxlxor 0, 0, 0
-; FMF-NEXT: fcmpu 0, 1, 0
-; FMF-NEXT: beq 0, .LBB16_2
-; FMF-NEXT: # %bb.1:
+; FMF-NEXT: # kill: def $f1 killed $f1 def $vsl1
; FMF-NEXT: xsrsqrtesp 0, 1
; FMF-NEXT: vspltisw 2, -3
; FMF-NEXT: addis 3, 2, .LCPI16_0 at toc@ha
-; FMF-NEXT: xvcvsxwdp 2, 34
-; FMF-NEXT: xsmulsp 1, 1, 0
-; FMF-NEXT: xsmaddasp 2, 1, 0
+; FMF-NEXT: xvcvsxwdp 3, 34
+; FMF-NEXT: xsmulsp 2, 1, 0
+; FMF-NEXT: xsmaddasp 3, 2, 0
; FMF-NEXT: lfs 0, .LCPI16_0 at toc@l(3)
-; FMF-NEXT: xsmulsp 0, 1, 0
-; FMF-NEXT: xsmulsp 0, 0, 2
-; FMF-NEXT: .LBB16_2:
-; FMF-NEXT: fmr 1, 0
+; FMF-NEXT: xsmulsp 0, 2, 0
+; FMF-NEXT: xxlxor 2, 2, 2
+; FMF-NEXT: xvcmpeqsp 1, 1, 2
+; FMF-NEXT: xsmulsp 0, 0, 3
+; FMF-NEXT: xxsel 1, 2, 0, 1
+; FMF-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; FMF-NEXT: blr
;
; GLOBAL-LABEL: sqrt_fast_preserve_sign:
; GLOBAL: # %bb.0:
-; GLOBAL-NEXT: xxlxor 0, 0, 0
-; GLOBAL-NEXT: fcmpu 0, 1, 0
-; GLOBAL-NEXT: beq 0, .LBB16_2
-; GLOBAL-NEXT: # %bb.1:
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 def $vsl1
; GLOBAL-NEXT: xsrsqrtesp 0, 1
; GLOBAL-NEXT: vspltisw 2, -3
; GLOBAL-NEXT: addis 3, 2, .LCPI16_0 at toc@ha
-; GLOBAL-NEXT: xvcvsxwdp 2, 34
-; GLOBAL-NEXT: xsmulsp 1, 1, 0
-; GLOBAL-NEXT: xsmaddasp 2, 1, 0
+; GLOBAL-NEXT: xvcvsxwdp 3, 34
+; GLOBAL-NEXT: xsmulsp 2, 1, 0
+; GLOBAL-NEXT: xsmaddasp 3, 2, 0
; GLOBAL-NEXT: lfs 0, .LCPI16_0 at toc@l(3)
-; GLOBAL-NEXT: xsmulsp 0, 1, 0
-; GLOBAL-NEXT: xsmulsp 0, 0, 2
-; GLOBAL-NEXT: .LBB16_2:
-; GLOBAL-NEXT: fmr 1, 0
+; GLOBAL-NEXT: xsmulsp 0, 2, 0
+; GLOBAL-NEXT: xxlxor 2, 2, 2
+; GLOBAL-NEXT: xvcmpeqsp 1, 1, 2
+; GLOBAL-NEXT: xsmulsp 0, 0, 3
+; GLOBAL-NEXT: xxsel 1, 2, 0, 1
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; GLOBAL-NEXT: blr
%rt = call contract reassoc ninf afn float @llvm.sqrt.f32(float %x)
ret float %rt
@@ -540,34 +528,34 @@ define float @sqrt_fast_preserve_sign(float %x) #1 {
; fcmp can have fast-math-flags.
; FMFDEBUG-LABEL: Optimized lowered selection DAG: %bb.0 'fcmp_nnan:'
-; FMFDEBUG: select_cc nnan {{t[0-9]+}}
+; FMFDEBUG: PPCISD::VSX_CMPSEL
; FMFDEBUG: Type-legalized selection DAG: %bb.0 'fcmp_nnan:'
; GLOBALDEBUG-LABEL: Optimized lowered selection DAG: %bb.0 'fcmp_nnan:'
-; GLOBALDEBUG: select_cc nnan {{t[0-9]+}}
+; GLOBALDEBUG: PPCISD::VSX_CMPSEL
; GLOBALDEBUG: Type-legalized selection DAG: %bb.0 'fcmp_nnan:'
define double @fcmp_nnan(double %a, double %y, double %z) {
; FMF-LABEL: fcmp_nnan:
; FMF: # %bb.0:
; FMF-NEXT: xxlxor 0, 0, 0
-; FMF-NEXT: xscmpudp 0, 1, 0
-; FMF-NEXT: blt 0, .LBB17_2
-; FMF-NEXT: # %bb.1:
-; FMF-NEXT: fmr 3, 2
-; FMF-NEXT: .LBB17_2:
-; FMF-NEXT: fmr 1, 3
+; FMF-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; FMF-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; FMF-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; FMF-NEXT: xvcmpgedp 0, 0, 1
+; FMF-NEXT: xxsel 1, 3, 2, 0
+; FMF-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; FMF-NEXT: blr
;
; GLOBAL-LABEL: fcmp_nnan:
; GLOBAL: # %bb.0:
; GLOBAL-NEXT: xxlxor 0, 0, 0
-; GLOBAL-NEXT: xscmpudp 0, 1, 0
-; GLOBAL-NEXT: blt 0, .LBB17_2
-; GLOBAL-NEXT: # %bb.1:
-; GLOBAL-NEXT: fmr 3, 2
-; GLOBAL-NEXT: .LBB17_2:
-; GLOBAL-NEXT: fmr 1, 3
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; GLOBAL-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; GLOBAL-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; GLOBAL-NEXT: xvcmpgedp 0, 0, 1
+; GLOBAL-NEXT: xxsel 1, 3, 2, 0
+; GLOBAL-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; GLOBAL-NEXT: blr
%cmp = fcmp nnan ult double %a, 0.0
%z.y = select i1 %cmp, double %z, double %y
diff --git a/llvm/test/CodeGen/PowerPC/recipest.ll b/llvm/test/CodeGen/PowerPC/recipest.ll
index 96e2c6c3e081a..8a0095d176b68 100644
--- a/llvm/test/CodeGen/PowerPC/recipest.ll
+++ b/llvm/test/CodeGen/PowerPC/recipest.ll
@@ -950,46 +950,42 @@ define float @goo3_fmf(float %a) nounwind {
;
; CHECK-P8-LABEL: goo3_fmf:
; CHECK-P8: # %bb.0:
-; CHECK-P8-NEXT: addis 3, 2, .LCPI23_1 at toc@ha
-; CHECK-P8-NEXT: xsabsdp 0, 1
-; CHECK-P8-NEXT: lfs 2, .LCPI23_1 at toc@l(3)
-; CHECK-P8-NEXT: fcmpu 0, 0, 2
-; CHECK-P8-NEXT: xxlxor 0, 0, 0
-; CHECK-P8-NEXT: blt 0, .LBB23_2
-; CHECK-P8-NEXT: # %bb.1:
; CHECK-P8-NEXT: xsrsqrtesp 0, 1
; CHECK-P8-NEXT: vspltisw 2, -3
; CHECK-P8-NEXT: addis 3, 2, .LCPI23_0 at toc@ha
-; CHECK-P8-NEXT: xvcvsxwdp 2, 34
-; CHECK-P8-NEXT: xsmulsp 1, 1, 0
-; CHECK-P8-NEXT: xsmaddasp 2, 1, 0
+; CHECK-P8-NEXT: xvcvsxwdp 3, 34
+; CHECK-P8-NEXT: xsmulsp 2, 1, 0
+; CHECK-P8-NEXT: xsabsdp 1, 1
+; CHECK-P8-NEXT: xsmaddasp 3, 2, 0
; CHECK-P8-NEXT: lfs 0, .LCPI23_0 at toc@l(3)
-; CHECK-P8-NEXT: xsmulsp 0, 1, 0
-; CHECK-P8-NEXT: xsmulsp 0, 0, 2
-; CHECK-P8-NEXT: .LBB23_2:
-; CHECK-P8-NEXT: fmr 1, 0
+; CHECK-P8-NEXT: addis 3, 2, .LCPI23_1 at toc@ha
+; CHECK-P8-NEXT: xsmulsp 0, 2, 0
+; CHECK-P8-NEXT: lfs 2, .LCPI23_1 at toc@l(3)
+; CHECK-P8-NEXT: xvcmpgesp 1, 2, 1
+; CHECK-P8-NEXT: xxlxor 2, 2, 2
+; CHECK-P8-NEXT: xsmulsp 0, 0, 3
+; CHECK-P8-NEXT: xxsel 1, 2, 0, 1
+; CHECK-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; CHECK-P8-NEXT: blr
;
; CHECK-P9-LABEL: goo3_fmf:
; CHECK-P9: # %bb.0:
-; CHECK-P9-NEXT: addis 3, 2, .LCPI23_1 at toc@ha
-; CHECK-P9-NEXT: xsabsdp 0, 1
-; CHECK-P9-NEXT: lfs 2, .LCPI23_1 at toc@l(3)
-; CHECK-P9-NEXT: fcmpu 0, 0, 2
-; CHECK-P9-NEXT: xxlxor 0, 0, 0
-; CHECK-P9-NEXT: blt 0, .LBB23_2
-; CHECK-P9-NEXT: # %bb.1:
; CHECK-P9-NEXT: xsrsqrtesp 0, 1
; CHECK-P9-NEXT: vspltisw 2, -3
; CHECK-P9-NEXT: addis 3, 2, .LCPI23_0 at toc@ha
-; CHECK-P9-NEXT: xsmulsp 1, 1, 0
-; CHECK-P9-NEXT: xvcvsxwdp 2, 34
-; CHECK-P9-NEXT: xsmaddasp 2, 1, 0
+; CHECK-P9-NEXT: xsmulsp 2, 1, 0
+; CHECK-P9-NEXT: xvcvsxwdp 3, 34
+; CHECK-P9-NEXT: xsabsdp 1, 1
+; CHECK-P9-NEXT: xsmaddasp 3, 2, 0
; CHECK-P9-NEXT: lfs 0, .LCPI23_0 at toc@l(3)
-; CHECK-P9-NEXT: xsmulsp 0, 1, 0
-; CHECK-P9-NEXT: xsmulsp 0, 0, 2
-; CHECK-P9-NEXT: .LBB23_2:
-; CHECK-P9-NEXT: fmr 1, 0
+; CHECK-P9-NEXT: addis 3, 2, .LCPI23_1 at toc@ha
+; CHECK-P9-NEXT: xsmulsp 0, 2, 0
+; CHECK-P9-NEXT: lfs 2, .LCPI23_1 at toc@l(3)
+; CHECK-P9-NEXT: xsmulsp 0, 0, 3
+; CHECK-P9-NEXT: xvcmpgesp 1, 2, 1
+; CHECK-P9-NEXT: xxlxor 2, 2, 2
+; CHECK-P9-NEXT: xxsel 1, 2, 0, 1
+; CHECK-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; CHECK-P9-NEXT: blr
%r = call contract reassoc ninf afn float @llvm.sqrt.f32(float %a)
ret float %r
diff --git a/llvm/test/CodeGen/PowerPC/scalar-equal.ll b/llvm/test/CodeGen/PowerPC/scalar-equal.ll
index 1832475e7795b..6d444d841a9fb 100644
--- a/llvm/test/CodeGen/PowerPC/scalar-equal.ll
+++ b/llvm/test/CodeGen/PowerPC/scalar-equal.ll
@@ -35,22 +35,22 @@ define double @testoeq(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P9-LABEL: testoeq:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P9-NEXT: beq cr0, .LBB0_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB0_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpeqdp vs0, f1, f2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
;
; NO-FAST-P8-LABEL: testoeq:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P8-NEXT: beq cr0, .LBB0_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB0_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpeqdp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
entry:
%cmp = fcmp oeq double %a, %b
diff --git a/llvm/test/CodeGen/PowerPC/scalar_cmp.ll b/llvm/test/CodeGen/PowerPC/scalar_cmp.ll
index aaabd76e163bb..184d7d6ef051c 100644
--- a/llvm/test/CodeGen/PowerPC/scalar_cmp.ll
+++ b/llvm/test/CodeGen/PowerPC/scalar_cmp.ll
@@ -36,22 +36,24 @@ define float @select_oeq_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_oeq_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: beq cr0, .LBB0_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB0_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpeqsp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_oeq_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: beq cr0, .LBB0_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB0_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpeqsp vs0, vs1, vs2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp oeq float %a, %b
@@ -78,22 +80,22 @@ define double @select_oeq_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_oeq_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P8-NEXT: beq cr0, .LBB1_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB1_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpeqdp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_oeq_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P9-NEXT: beq cr0, .LBB1_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB1_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpeqdp vs0, f1, f2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp oeq double %a, %b
@@ -198,24 +200,24 @@ define float @select_one_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_one_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, eq
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB4_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB4_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpeqsp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_one_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, eq
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB4_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB4_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpeqsp vs0, vs2, vs1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp one float %a, %b
@@ -242,24 +244,22 @@ define double @select_one_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_one_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, eq
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB5_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB5_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpeqdp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_one_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, eq
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB5_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB5_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpeqdp vs0, f2, f1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp one double %a, %b
@@ -360,24 +360,24 @@ define float @select_oge_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_oge_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, lt
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB8_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB8_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgesp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_oge_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, lt
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB8_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB8_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpgesp vs0, vs1, vs2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp oge float %a, %b
@@ -400,24 +400,22 @@ define double @select_oge_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_oge_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, lt
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB9_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB9_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgedp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_oge_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, lt
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB9_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB9_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpgedp vs0, f1, f2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp oge double %a, %b
@@ -502,22 +500,24 @@ define float @select_olt_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_olt_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: blt cr0, .LBB12_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB12_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgesp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_olt_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: blt cr0, .LBB12_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB12_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpgesp vs0, vs2, vs1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp olt float %a, %b
@@ -540,22 +540,22 @@ define double @select_olt_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_olt_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P8-NEXT: blt cr0, .LBB13_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB13_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgedp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_olt_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P9-NEXT: blt cr0, .LBB13_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB13_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpgedp vs0, f2, f1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp olt double %a, %b
@@ -640,22 +640,24 @@ define float @select_ogt_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_ogt_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: bgt cr0, .LBB16_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB16_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgtsp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_ogt_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: bgt cr0, .LBB16_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB16_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpgtsp vs0, vs1, vs2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp ogt float %a, %b
@@ -678,22 +680,22 @@ define double @select_ogt_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_ogt_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P8-NEXT: bgt cr0, .LBB17_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB17_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgtdp vs0, vs1, vs2
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_ogt_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f2
-; NO-FAST-P9-NEXT: bgt cr0, .LBB17_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB17_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpgtdp vs0, f1, f2
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp ogt double %a, %b
@@ -778,24 +780,24 @@ define float @select_ole_float(float %a, float %b, float %c, float %d) {
;
; NO-FAST-P8-LABEL: select_ole_float:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, gt
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB20_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB20_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgtsp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_ole_float:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, gt
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB20_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB20_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P9-NEXT: xvcmpgtsp vs0, vs2, vs1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp ole float %a, %b
@@ -818,24 +820,22 @@ define double @select_ole_double(double %a, double %b, double %c, double %d) {
;
; NO-FAST-P8-LABEL: select_ole_double:
; NO-FAST-P8: # %bb.0: # %entry
-; NO-FAST-P8-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P8-NEXT: crnor 4*cr5+lt, un, gt
-; NO-FAST-P8-NEXT: bc 12, 4*cr5+lt, .LBB21_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f3, f4
-; NO-FAST-P8-NEXT: .LBB21_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: xvcmpgtdp vs0, vs2, vs1
+; NO-FAST-P8-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: select_ole_double:
; NO-FAST-P9: # %bb.0: # %entry
-; NO-FAST-P9-NEXT: fcmpu cr0, f1, f2
-; NO-FAST-P9-NEXT: crnor 4*cr5+lt, un, gt
-; NO-FAST-P9-NEXT: bc 12, 4*cr5+lt, .LBB21_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f3, f4
-; NO-FAST-P9-NEXT: .LBB21_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f3
+; NO-FAST-P9-NEXT: xscmpgtdp vs0, f2, f1
+; NO-FAST-P9-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: xxsel vs1, vs3, vs4, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp ole double %a, %b
@@ -976,25 +976,24 @@ define double @onecmp2(double %a, double %y, double %z) {
; NO-FAST-P8-LABEL: onecmp2:
; NO-FAST-P8: # %bb.0: # %entry
; NO-FAST-P8-NEXT: vspltisw v2, 1
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
; NO-FAST-P8-NEXT: xvcvsxwdp vs0, vs34
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f0
-; NO-FAST-P8-NEXT: bgt cr0, .LBB25_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f2, f3
-; NO-FAST-P8-NEXT: .LBB25_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f2
+; NO-FAST-P8-NEXT: xvcmpgtdp vs0, vs1, vs0
+; NO-FAST-P8-NEXT: xxsel vs1, vs2, vs3, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: onecmp2:
; NO-FAST-P9: # %bb.0: # %entry
; NO-FAST-P9-NEXT: vspltisw v2, 1
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
; NO-FAST-P9-NEXT: xvcvsxwdp vs0, vs34
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f0
-; NO-FAST-P9-NEXT: bgt cr0, .LBB25_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f2, f3
-; NO-FAST-P9-NEXT: .LBB25_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f2
+; NO-FAST-P9-NEXT: xscmpgtdp vs0, f1, f0
+; NO-FAST-P9-NEXT: xxsel vs1, vs2, vs3, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp ogt double %a, 1.000000e+00
@@ -1026,25 +1025,24 @@ define double @onecmp3(double %a, double %y, double %z) {
; NO-FAST-P8-LABEL: onecmp3:
; NO-FAST-P8: # %bb.0: # %entry
; NO-FAST-P8-NEXT: vspltisw v2, 1
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; NO-FAST-P8-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P8-NEXT: # kill: def $f2 killed $f2 def $vsl2
; NO-FAST-P8-NEXT: xvcvsxwdp vs0, vs34
-; NO-FAST-P8-NEXT: xscmpudp cr0, f1, f0
-; NO-FAST-P8-NEXT: beq cr0, .LBB26_2
-; NO-FAST-P8-NEXT: # %bb.1: # %entry
-; NO-FAST-P8-NEXT: fmr f2, f3
-; NO-FAST-P8-NEXT: .LBB26_2: # %entry
-; NO-FAST-P8-NEXT: fmr f1, f2
+; NO-FAST-P8-NEXT: xvcmpeqdp vs0, vs1, vs0
+; NO-FAST-P8-NEXT: xxsel vs1, vs2, vs3, vs0
+; NO-FAST-P8-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P8-NEXT: blr
;
; NO-FAST-P9-LABEL: onecmp3:
; NO-FAST-P9: # %bb.0: # %entry
; NO-FAST-P9-NEXT: vspltisw v2, 1
+; NO-FAST-P9-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; NO-FAST-P9-NEXT: # kill: def $f2 killed $f2 def $vsl2
; NO-FAST-P9-NEXT: xvcvsxwdp vs0, vs34
-; NO-FAST-P9-NEXT: xscmpudp cr0, f1, f0
-; NO-FAST-P9-NEXT: beq cr0, .LBB26_2
-; NO-FAST-P9-NEXT: # %bb.1: # %entry
-; NO-FAST-P9-NEXT: fmr f2, f3
-; NO-FAST-P9-NEXT: .LBB26_2: # %entry
-; NO-FAST-P9-NEXT: fmr f1, f2
+; NO-FAST-P9-NEXT: xscmpeqdp vs0, f1, f0
+; NO-FAST-P9-NEXT: xxsel vs1, vs2, vs3, vs0
+; NO-FAST-P9-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; NO-FAST-P9-NEXT: blr
entry:
%cmp = fcmp oeq double %a, 1.000000e+00
diff --git a/llvm/test/CodeGen/PowerPC/vsx.ll b/llvm/test/CodeGen/PowerPC/vsx.ll
index 32cbfd6d810ac..5432476ccf30e 100644
--- a/llvm/test/CodeGen/PowerPC/vsx.ll
+++ b/llvm/test/CodeGen/PowerPC/vsx.ll
@@ -2465,18 +2465,24 @@ define <2 x double> @test81(<4 x float> %b) {
define double @test82(double %a, double %b, double %c, double %d) {
; CHECK-LABEL: test82:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: xscmpudp cr0, f3, f4
-; CHECK-NEXT: beqlr cr0
-; CHECK-NEXT: # %bb.1: # %entry
-; CHECK-NEXT: fmr f1, f2
+; CHECK-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; CHECK-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; CHECK-NEXT: xvcmpeqdp vs0, vs3, vs4
+; CHECK-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; CHECK-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; CHECK-NEXT: xxsel vs1, vs1, vs2, vs0
+; CHECK-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; CHECK-NEXT: blr
;
; CHECK-REG-LABEL: test82:
; CHECK-REG: # %bb.0: # %entry
-; CHECK-REG-NEXT: xscmpudp cr0, f3, f4
-; CHECK-REG-NEXT: beqlr cr0
-; CHECK-REG-NEXT: # %bb.1: # %entry
-; CHECK-REG-NEXT: fmr f1, f2
+; CHECK-REG-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; CHECK-REG-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; CHECK-REG-NEXT: xvcmpeqdp vs0, vs3, vs4
+; CHECK-REG-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; CHECK-REG-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; CHECK-REG-NEXT: xxsel vs1, vs1, vs2, vs0
+; CHECK-REG-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; CHECK-REG-NEXT: blr
;
; CHECK-FISL-LABEL: test82:
@@ -2495,10 +2501,13 @@ define double @test82(double %a, double %b, double %c, double %d) {
;
; CHECK-LE-LABEL: test82:
; CHECK-LE: # %bb.0: # %entry
-; CHECK-LE-NEXT: xscmpudp cr0, f3, f4
-; CHECK-LE-NEXT: beqlr cr0
-; CHECK-LE-NEXT: # %bb.1: # %entry
-; CHECK-LE-NEXT: fmr f1, f2
+; CHECK-LE-NEXT: # kill: def $f4 killed $f4 def $vsl4
+; CHECK-LE-NEXT: # kill: def $f3 killed $f3 def $vsl3
+; CHECK-LE-NEXT: xvcmpeqdp vs0, vs3, vs4
+; CHECK-LE-NEXT: # kill: def $f2 killed $f2 def $vsl2
+; CHECK-LE-NEXT: # kill: def $f1 killed $f1 def $vsl1
+; CHECK-LE-NEXT: xxsel vs1, vs1, vs2, vs0
+; CHECK-LE-NEXT: # kill: def $f1 killed $f1 killed $vsl1
; CHECK-LE-NEXT: blr
entry:
%m = fcmp oeq double %c, %d
More information about the llvm-commits
mailing list