cmd/compile: generate subfic on ppc64

This merges an lis + subf into subfic, and for 32b constants
lwa + subf into oris + ori + subf.

The carry bit is no longer used in code generation, therefore
I think we can clobber it as needed.  Note, lowered borrow/carry
arithmetic is self-contained and thus is not affected.

A few extra rules are added to ensure early transformations to
SUBFCconst don't trip up earlier rules, fold constant operations,
or otherwise simplify lowering.  Likewise, tests are added to
ensure all rules are hit.  Generic constant folding catches
trivial cases, however some lowering rules insert arithmetic
which can introduce new opportunities (e.g BitLen or Slicemask).

I couldn't find a specific benchmark to demonstrate noteworthy
improvements, but this is generating subfic in many of the default
bent test binaries, so we are at least saving a little code space.

Change-Id: Iad7c6e5767eaa9dc24dc1c989bd1c8cfe1982012
Reviewed-on: https://go-review.googlesource.com/c/go/+/249461
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com>

1 files changed, 557 insertions, 14 deletions

diff --git a/src/cmd/compile/internal/ssa/rewritePPC64.go b/src/cmd/compile/internal/ssa/rewritePPC64.go
index 1a0b03e81c..152cdfdf4d 100644
--- a/src/cmd/compile/internal/ssa/rewritePPC64.go
+++ b/src/cmd/compile/internal/ssa/rewritePPC64.go
@@ -568,6 +568,8 @@ func rewriteValuePPC64(v *Value) bool {
 		return rewriteValuePPC64_OpPPC64MOVWstorezero(v)
 	case OpPPC64MTVSRD:
 		return rewriteValuePPC64_OpPPC64MTVSRD(v)
+	case OpPPC64NEG:
+		return rewriteValuePPC64_OpPPC64NEG(v)
 	case OpPPC64NOR:
 		return rewriteValuePPC64_OpPPC64NOR(v)
 	case OpPPC64NotEqual:
@@ -596,6 +598,8 @@ func rewriteValuePPC64(v *Value) bool {
 		return rewriteValuePPC64_OpPPC64SRW(v)
 	case OpPPC64SUB:
 		return rewriteValuePPC64_OpPPC64SUB(v)
+	case OpPPC64SUBFCconst:
+		return rewriteValuePPC64_OpPPC64SUBFCconst(v)
 	case OpPPC64XOR:
 		return rewriteValuePPC64_OpPPC64XOR(v)
 	case OpPPC64XORconst:
@@ -1021,15 +1025,14 @@ func rewriteValuePPC64_OpBitLen32(v *Value) bool {
 	b := v.Block
 	typ := &b.Func.Config.Types
 	// match: (BitLen32 x)
-	// result: (SUB (MOVDconst [32]) (CNTLZW <typ.Int> x))
+	// result: (SUBFCconst [32] (CNTLZW <typ.Int> x))
 	for {
 		x := v_0
-		v.reset(OpPPC64SUB)
-		v0 := b.NewValue0(v.Pos, OpPPC64MOVDconst, typ.Int64)
-		v0.AuxInt = int64ToAuxInt(32)
-		v1 := b.NewValue0(v.Pos, OpPPC64CNTLZW, typ.Int)
-		v1.AddArg(x)
-		v.AddArg2(v0, v1)
+		v.reset(OpPPC64SUBFCconst)
+		v.AuxInt = int64ToAuxInt(32)
+		v0 := b.NewValue0(v.Pos, OpPPC64CNTLZW, typ.Int)
+		v0.AddArg(x)
+		v.AddArg(v0)
 		return true
 	}
 }
@@ -1038,15 +1041,14 @@ func rewriteValuePPC64_OpBitLen64(v *Value) bool {
 	b := v.Block
 	typ := &b.Func.Config.Types
 	// match: (BitLen64 x)
-	// result: (SUB (MOVDconst [64]) (CNTLZD <typ.Int> x))
+	// result: (SUBFCconst [64] (CNTLZD <typ.Int> x))
 	for {
 		x := v_0
-		v.reset(OpPPC64SUB)
-		v0 := b.NewValue0(v.Pos, OpPPC64MOVDconst, typ.Int64)
-		v0.AuxInt = int64ToAuxInt(64)
-		v1 := b.NewValue0(v.Pos, OpPPC64CNTLZD, typ.Int)
-		v1.AddArg(x)
-		v.AddArg2(v0, v1)
+		v.reset(OpPPC64SUBFCconst)
+		v.AuxInt = int64ToAuxInt(64)
+		v0 := b.NewValue0(v.Pos, OpPPC64CNTLZD, typ.Int)
+		v0.AddArg(x)
+		v.AddArg(v0)
 		return true
 	}
 }
@@ -3957,6 +3959,76 @@ func rewriteValuePPC64_OpPPC64ADD(v *Value) bool {
 		}
 		break
 	}
+	// match: (ADD (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y))))
+	// result: (ROTL x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLD {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int64 || auxIntToInt64(v_0_1.AuxInt) != 63 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRD {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 64 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 63 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTL)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
+	// match: (ADD (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y))))
+	// result: (ROTLW x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLW {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int32 || auxIntToInt64(v_0_1.AuxInt) != 31 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRW {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 32 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 31 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTLW)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
 	// match: (ADD (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y))))
 	// result: (ROTLW x y)
 	for {
@@ -4069,6 +4141,24 @@ func rewriteValuePPC64_OpPPC64ADDconst(v *Value) bool {
 		v.AddArg(x)
 		return true
 	}
+	// match: (ADDconst [c] (SUBFCconst [d] x))
+	// cond: is32Bit(c+d)
+	// result: (SUBFCconst [c+d] x)
+	for {
+		c := auxIntToInt64(v.AuxInt)
+		if v_0.Op != OpPPC64SUBFCconst {
+			break
+		}
+		d := auxIntToInt64(v_0.AuxInt)
+		x := v_0.Args[0]
+		if !(is32Bit(c + d)) {
+			break
+		}
+		v.reset(OpPPC64SUBFCconst)
+		v.AuxInt = int64ToAuxInt(c + d)
+		v.AddArg(x)
+		return true
+	}
 	return false
 }
 func rewriteValuePPC64_OpPPC64AND(v *Value) bool {
@@ -10336,6 +10426,44 @@ func rewriteValuePPC64_OpPPC64MTVSRD(v *Value) bool {
 	}
 	return false
 }
+func rewriteValuePPC64_OpPPC64NEG(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (NEG (ADDconst [c] x))
+	// cond: is32Bit(-c)
+	// result: (SUBFCconst [-c] x)
+	for {
+		if v_0.Op != OpPPC64ADDconst {
+			break
+		}
+		c := auxIntToInt64(v_0.AuxInt)
+		x := v_0.Args[0]
+		if !(is32Bit(-c)) {
+			break
+		}
+		v.reset(OpPPC64SUBFCconst)
+		v.AuxInt = int64ToAuxInt(-c)
+		v.AddArg(x)
+		return true
+	}
+	// match: (NEG (SUBFCconst [c] x))
+	// cond: is32Bit(-c)
+	// result: (ADDconst [-c] x)
+	for {
+		if v_0.Op != OpPPC64SUBFCconst {
+			break
+		}
+		c := auxIntToInt64(v_0.AuxInt)
+		x := v_0.Args[0]
+		if !(is32Bit(-c)) {
+			break
+		}
+		v.reset(OpPPC64ADDconst)
+		v.AuxInt = int64ToAuxInt(-c)
+		v.AddArg(x)
+		return true
+	}
+	return false
+}
 func rewriteValuePPC64_OpPPC64NOR(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
@@ -10510,6 +10638,76 @@ func rewriteValuePPC64_OpPPC64OR(v *Value) bool {
 		}
 		break
 	}
+	// match: ( OR (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y))))
+	// result: (ROTL x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLD {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int64 || auxIntToInt64(v_0_1.AuxInt) != 63 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRD {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 64 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 63 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTL)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
+	// match: ( OR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y))))
+	// result: (ROTLW x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLW {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int32 || auxIntToInt64(v_0_1.AuxInt) != 31 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRW {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 32 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 31 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTLW)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
 	// match: ( OR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y))))
 	// result: (ROTLW x y)
 	for {
@@ -12109,6 +12307,69 @@ func rewriteValuePPC64_OpPPC64SUB(v *Value) bool {
 		v.AddArg(x)
 		return true
 	}
+	// match: (SUB (MOVDconst [c]) x)
+	// cond: is32Bit(c)
+	// result: (SUBFCconst [c] x)
+	for {
+		if v_0.Op != OpPPC64MOVDconst {
+			break
+		}
+		c := auxIntToInt64(v_0.AuxInt)
+		x := v_1
+		if !(is32Bit(c)) {
+			break
+		}
+		v.reset(OpPPC64SUBFCconst)
+		v.AuxInt = int64ToAuxInt(c)
+		v.AddArg(x)
+		return true
+	}
+	return false
+}
+func rewriteValuePPC64_OpPPC64SUBFCconst(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (SUBFCconst [c] (NEG x))
+	// result: (ADDconst [c] x)
+	for {
+		c := auxIntToInt64(v.AuxInt)
+		if v_0.Op != OpPPC64NEG {
+			break
+		}
+		x := v_0.Args[0]
+		v.reset(OpPPC64ADDconst)
+		v.AuxInt = int64ToAuxInt(c)
+		v.AddArg(x)
+		return true
+	}
+	// match: (SUBFCconst [c] (SUBFCconst [d] x))
+	// cond: is32Bit(c-d)
+	// result: (ADDconst [c-d] x)
+	for {
+		c := auxIntToInt64(v.AuxInt)
+		if v_0.Op != OpPPC64SUBFCconst {
+			break
+		}
+		d := auxIntToInt64(v_0.AuxInt)
+		x := v_0.Args[0]
+		if !(is32Bit(c - d)) {
+			break
+		}
+		v.reset(OpPPC64ADDconst)
+		v.AuxInt = int64ToAuxInt(c - d)
+		v.AddArg(x)
+		return true
+	}
+	// match: (SUBFCconst [0] x)
+	// result: (NEG x)
+	for {
+		if auxIntToInt64(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.reset(OpPPC64NEG)
+		v.AddArg(x)
+		return true
+	}
 	return false
 }
 func rewriteValuePPC64_OpPPC64XOR(v *Value) bool {
@@ -12204,6 +12465,76 @@ func rewriteValuePPC64_OpPPC64XOR(v *Value) bool {
 		}
 		break
 	}
+	// match: (XOR (SLD x (ANDconst <typ.Int64> [63] y)) (SRD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y))))
+	// result: (ROTL x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLD {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int64 || auxIntToInt64(v_0_1.AuxInt) != 63 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRD {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 64 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 63 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTL)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
+	// match: (XOR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y))))
+	// result: (ROTLW x y)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			if v_0.Op != OpPPC64SLW {
+				continue
+			}
+			_ = v_0.Args[1]
+			x := v_0.Args[0]
+			v_0_1 := v_0.Args[1]
+			if v_0_1.Op != OpPPC64ANDconst || v_0_1.Type != typ.Int32 || auxIntToInt64(v_0_1.AuxInt) != 31 {
+				continue
+			}
+			y := v_0_1.Args[0]
+			if v_1.Op != OpPPC64SRW {
+				continue
+			}
+			_ = v_1.Args[1]
+			if x != v_1.Args[0] {
+				continue
+			}
+			v_1_1 := v_1.Args[1]
+			if v_1_1.Op != OpPPC64SUBFCconst || v_1_1.Type != typ.UInt || auxIntToInt64(v_1_1.AuxInt) != 32 {
+				continue
+			}
+			v_1_1_0 := v_1_1.Args[0]
+			if v_1_1_0.Op != OpPPC64ANDconst || v_1_1_0.Type != typ.UInt || auxIntToInt64(v_1_1_0.AuxInt) != 31 || y != v_1_1_0.Args[0] {
+				continue
+			}
+			v.reset(OpPPC64ROTLW)
+			v.AddArg2(x, y)
+			return true
+		}
+		break
+	}
 	// match: (XOR (SLW x (ANDconst <typ.Int32> [31] y)) (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y))))
 	// result: (ROTLW x y)
 	for {
@@ -13175,6 +13506,28 @@ func rewriteValuePPC64_OpRsh32Ux64(v *Value) bool {
 		v.AddArg2(x, v0)
 		return true
 	}
+	// match: (Rsh32Ux64 x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	// result: (SRW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 32 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64ANDconst || v_1_0.Type != typ.UInt || auxIntToInt64(v_1_0.AuxInt) != 31 {
+			break
+		}
+		y := v_1_0.Args[0]
+		v.reset(OpPPC64SRW)
+		v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+		v0.AuxInt = int64ToAuxInt(32)
+		v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+		v1.AuxInt = int64ToAuxInt(31)
+		v1.AddArg(y)
+		v0.AddArg(v1)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Rsh32Ux64 x (SUB <typ.UInt> (MOVDconst [32]) (AND <typ.UInt> y (MOVDconst [31]))))
 	// result: (SRW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))
 	for {
@@ -13212,6 +13565,37 @@ func rewriteValuePPC64_OpRsh32Ux64(v *Value) bool {
 		}
 		break
 	}
+	// match: (Rsh32Ux64 x (SUBFCconst <typ.UInt> [32] (AND <typ.UInt> y (MOVDconst [31]))))
+	// result: (SRW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 32 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64AND || v_1_0.Type != typ.UInt {
+			break
+		}
+		_ = v_1_0.Args[1]
+		v_1_0_0 := v_1_0.Args[0]
+		v_1_0_1 := v_1_0.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_1_0_0, v_1_0_1 = _i0+1, v_1_0_1, v_1_0_0 {
+			y := v_1_0_0
+			if v_1_0_1.Op != OpPPC64MOVDconst || auxIntToInt64(v_1_0_1.AuxInt) != 31 {
+				continue
+			}
+			v.reset(OpPPC64SRW)
+			v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+			v0.AuxInt = int64ToAuxInt(32)
+			v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+			v1.AuxInt = int64ToAuxInt(31)
+			v1.AddArg(y)
+			v0.AddArg(v1)
+			v.AddArg2(x, v0)
+			return true
+		}
+		break
+	}
 	// match: (Rsh32Ux64 x y)
 	// result: (SRW x (ISEL [0] y (MOVDconst [-1]) (CMPU y (MOVDconst [32]))))
 	for {
@@ -13482,6 +13866,28 @@ func rewriteValuePPC64_OpRsh32x64(v *Value) bool {
 		v.AddArg2(x, v0)
 		return true
 	}
+	// match: (Rsh32x64 x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	// result: (SRAW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 32 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64ANDconst || v_1_0.Type != typ.UInt || auxIntToInt64(v_1_0.AuxInt) != 31 {
+			break
+		}
+		y := v_1_0.Args[0]
+		v.reset(OpPPC64SRAW)
+		v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+		v0.AuxInt = int64ToAuxInt(32)
+		v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+		v1.AuxInt = int64ToAuxInt(31)
+		v1.AddArg(y)
+		v0.AddArg(v1)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Rsh32x64 x (SUB <typ.UInt> (MOVDconst [32]) (AND <typ.UInt> y (MOVDconst [31]))))
 	// result: (SRAW x (SUB <typ.UInt> (MOVDconst [32]) (ANDconst <typ.UInt> [31] y)))
 	for {
@@ -13519,6 +13925,37 @@ func rewriteValuePPC64_OpRsh32x64(v *Value) bool {
 		}
 		break
 	}
+	// match: (Rsh32x64 x (SUBFCconst <typ.UInt> [32] (AND <typ.UInt> y (MOVDconst [31]))))
+	// result: (SRAW x (SUBFCconst <typ.UInt> [32] (ANDconst <typ.UInt> [31] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 32 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64AND || v_1_0.Type != typ.UInt {
+			break
+		}
+		_ = v_1_0.Args[1]
+		v_1_0_0 := v_1_0.Args[0]
+		v_1_0_1 := v_1_0.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_1_0_0, v_1_0_1 = _i0+1, v_1_0_1, v_1_0_0 {
+			y := v_1_0_0
+			if v_1_0_1.Op != OpPPC64MOVDconst || auxIntToInt64(v_1_0_1.AuxInt) != 31 {
+				continue
+			}
+			v.reset(OpPPC64SRAW)
+			v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+			v0.AuxInt = int64ToAuxInt(32)
+			v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+			v1.AuxInt = int64ToAuxInt(31)
+			v1.AddArg(y)
+			v0.AddArg(v1)
+			v.AddArg2(x, v0)
+			return true
+		}
+		break
+	}
 	// match: (Rsh32x64 x y)
 	// result: (SRAW x (ISEL [0] y (MOVDconst [-1]) (CMPU y (MOVDconst [32]))))
 	for {
@@ -13787,6 +14224,28 @@ func rewriteValuePPC64_OpRsh64Ux64(v *Value) bool {
 		v.AddArg2(x, v0)
 		return true
 	}
+	// match: (Rsh64Ux64 x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	// result: (SRD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 64 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64ANDconst || v_1_0.Type != typ.UInt || auxIntToInt64(v_1_0.AuxInt) != 63 {
+			break
+		}
+		y := v_1_0.Args[0]
+		v.reset(OpPPC64SRD)
+		v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+		v0.AuxInt = int64ToAuxInt(64)
+		v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+		v1.AuxInt = int64ToAuxInt(63)
+		v1.AddArg(y)
+		v0.AddArg(v1)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Rsh64Ux64 x (SUB <typ.UInt> (MOVDconst [64]) (AND <typ.UInt> y (MOVDconst [63]))))
 	// result: (SRD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))
 	for {
@@ -13824,6 +14283,37 @@ func rewriteValuePPC64_OpRsh64Ux64(v *Value) bool {
 		}
 		break
 	}
+	// match: (Rsh64Ux64 x (SUBFCconst <typ.UInt> [64] (AND <typ.UInt> y (MOVDconst [63]))))
+	// result: (SRD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 64 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64AND || v_1_0.Type != typ.UInt {
+			break
+		}
+		_ = v_1_0.Args[1]
+		v_1_0_0 := v_1_0.Args[0]
+		v_1_0_1 := v_1_0.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_1_0_0, v_1_0_1 = _i0+1, v_1_0_1, v_1_0_0 {
+			y := v_1_0_0
+			if v_1_0_1.Op != OpPPC64MOVDconst || auxIntToInt64(v_1_0_1.AuxInt) != 63 {
+				continue
+			}
+			v.reset(OpPPC64SRD)
+			v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+			v0.AuxInt = int64ToAuxInt(64)
+			v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+			v1.AuxInt = int64ToAuxInt(63)
+			v1.AddArg(y)
+			v0.AddArg(v1)
+			v.AddArg2(x, v0)
+			return true
+		}
+		break
+	}
 	// match: (Rsh64Ux64 x y)
 	// result: (SRD x (ISEL [0] y (MOVDconst [-1]) (CMPU y (MOVDconst [64]))))
 	for {
@@ -14094,6 +14584,28 @@ func rewriteValuePPC64_OpRsh64x64(v *Value) bool {
 		v.AddArg2(x, v0)
 		return true
 	}
+	// match: (Rsh64x64 x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	// result: (SRAD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 64 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64ANDconst || v_1_0.Type != typ.UInt || auxIntToInt64(v_1_0.AuxInt) != 63 {
+			break
+		}
+		y := v_1_0.Args[0]
+		v.reset(OpPPC64SRAD)
+		v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+		v0.AuxInt = int64ToAuxInt(64)
+		v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+		v1.AuxInt = int64ToAuxInt(63)
+		v1.AddArg(y)
+		v0.AddArg(v1)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Rsh64x64 x (SUB <typ.UInt> (MOVDconst [64]) (AND <typ.UInt> y (MOVDconst [63]))))
 	// result: (SRAD x (SUB <typ.UInt> (MOVDconst [64]) (ANDconst <typ.UInt> [63] y)))
 	for {
@@ -14131,6 +14643,37 @@ func rewriteValuePPC64_OpRsh64x64(v *Value) bool {
 		}
 		break
 	}
+	// match: (Rsh64x64 x (SUBFCconst <typ.UInt> [64] (AND <typ.UInt> y (MOVDconst [63]))))
+	// result: (SRAD x (SUBFCconst <typ.UInt> [64] (ANDconst <typ.UInt> [63] y)))
+	for {
+		x := v_0
+		if v_1.Op != OpPPC64SUBFCconst || v_1.Type != typ.UInt || auxIntToInt64(v_1.AuxInt) != 64 {
+			break
+		}
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpPPC64AND || v_1_0.Type != typ.UInt {
+			break
+		}
+		_ = v_1_0.Args[1]
+		v_1_0_0 := v_1_0.Args[0]
+		v_1_0_1 := v_1_0.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_1_0_0, v_1_0_1 = _i0+1, v_1_0_1, v_1_0_0 {
+			y := v_1_0_0
+			if v_1_0_1.Op != OpPPC64MOVDconst || auxIntToInt64(v_1_0_1.AuxInt) != 63 {
+				continue
+			}
+			v.reset(OpPPC64SRAD)
+			v0 := b.NewValue0(v.Pos, OpPPC64SUBFCconst, typ.UInt)
+			v0.AuxInt = int64ToAuxInt(64)
+			v1 := b.NewValue0(v.Pos, OpPPC64ANDconst, typ.UInt)
+			v1.AuxInt = int64ToAuxInt(63)
+			v1.AddArg(y)
+			v0.AddArg(v1)
+			v.AddArg2(x, v0)
+			return true
+		}
+		break
+	}
 	// match: (Rsh64x64 x y)
 	// result: (SRAD x (ISEL [0] y (MOVDconst [-1]) (CMPU y (MOVDconst [64]))))
 	for {