cmd/compile: ppc64 fold (x+x)<<c into x<<(c+1)

On ppc64/ppc64le, rewrite (x + x) << c to x << (c+1) for constant shifts. This removes an ADD, shortens the dependency chain, and reduces code size.

Add rules for both 64-bit (SLDconst) and 32-bit (SLWconst), and extend
test/codegen/shift.go with ppc64x checks to assert a single SLD/SLW and
forbid ADD. Aligns ppc64 with other architectures that already assert
similar codegen in shift.go.

Change-Id: Ie564afbb029a5bd48887b82b0c455ca1dddd5508
Cq-Include-Trybots: luci.golang.try:gotip-linux-ppc64_power10,gotip-linux-ppc64_power8,gotip-linux-ppc64le_power8,gotip-linux-ppc64le_power9,gotip-linux-ppc64le_power10
Reviewed-on: https://go-review.googlesource.com/c/go/+/712000
Reviewed-by: Archana Ravindar <aravinda@redhat.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>

2 files changed, 38 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/ssa/_gen/PPC64.rules b/src/cmd/compile/internal/ssa/_gen/PPC64.rules
index 3678a7e941..6d40687264 100644
--- a/src/cmd/compile/internal/ssa/_gen/PPC64.rules
+++ b/src/cmd/compile/internal/ssa/_gen/PPC64.rules
@@ -869,10 +869,14 @@
 
 (SLDconst [c] z:(ANDconst [d] x)) && z.Uses == 1 && isPPC64ValidShiftMask(d) && c <= (64-getPPC64ShiftMaskLength(d)) => (CLRLSLDI [newPPC64ShiftAuxInt(c,64-getPPC64ShiftMaskLength(d),63,64)] x)
 (SLDconst [c] z:(AND (MOVDconst [d]) x)) && z.Uses == 1 && isPPC64ValidShiftMask(d) && c<=(64-getPPC64ShiftMaskLength(d)) => (CLRLSLDI [newPPC64ShiftAuxInt(c,64-getPPC64ShiftMaskLength(d),63,64)] x)
+// (x + x) << c  →  x << (c+1)
+(SLDconst [c] (ADD x x)) && c < 63 => (SLDconst [c+1] x)
 (SLWconst [c] z:(MOVBZreg x)) && z.Uses == 1 && c < 8 => (CLRLSLWI [newPPC64ShiftAuxInt(c,24,31,32)] x)
 (SLWconst [c] z:(MOVHZreg x)) && z.Uses == 1 && c < 16 => (CLRLSLWI [newPPC64ShiftAuxInt(c,16,31,32)] x)
 (SLWconst [c] z:(ANDconst [d] x)) && z.Uses == 1 && isPPC64ValidShiftMask(d) && c<=(32-getPPC64ShiftMaskLength(d)) => (CLRLSLWI [newPPC64ShiftAuxInt(c,32-getPPC64ShiftMaskLength(d),31,32)] x)
 (SLWconst [c] z:(AND (MOVDconst [d]) x)) && z.Uses == 1 && isPPC64ValidShiftMask(d) && c<=(32-getPPC64ShiftMaskLength(d)) => (CLRLSLWI [newPPC64ShiftAuxInt(c,32-getPPC64ShiftMaskLength(d),31,32)] x)
+// (x + x) << c  →  x << (c+1)
+(SLWconst [c] (ADD x x)) && c < 31 => (SLWconst [c+1] x)
 // special case for power9
 (SL(W|D)const [c] z:(MOVWreg x)) && c < 32 && buildcfg.GOPPC64 >= 9 => (EXTSWSLconst [c] x)
 
diff --git a/src/cmd/compile/internal/ssa/rewritePPC64.go b/src/cmd/compile/internal/ssa/rewritePPC64.go
index 17a5421ff8..6a7df42546 100644
--- a/src/cmd/compile/internal/ssa/rewritePPC64.go
+++ b/src/cmd/compile/internal/ssa/rewritePPC64.go
@@ -12614,6 +12614,23 @@ func rewriteValuePPC64_OpPPC64SLDconst(v *Value) bool {
 		}
 		break
 	}
+	// match: (SLDconst [c] (ADD x x))
+	// cond: c < 63
+	// result: (SLDconst [c+1] x)
+	for {
+		c := auxIntToInt64(v.AuxInt)
+		if v_0.Op != OpPPC64ADD {
+			break
+		}
+		x := v_0.Args[1]
+		if x != v_0.Args[0] || !(c < 63) {
+			break
+		}
+		v.reset(OpPPC64SLDconst)
+		v.AuxInt = int64ToAuxInt(c + 1)
+		v.AddArg(x)
+		return true
+	}
 	// match: (SLDconst [c] z:(MOVWreg x))
 	// cond: c < 32 && buildcfg.GOPPC64 >= 9
 	// result: (EXTSWSLconst [c] x)
@@ -12750,6 +12767,23 @@ func rewriteValuePPC64_OpPPC64SLWconst(v *Value) bool {
 		}
 		break
 	}
+	// match: (SLWconst [c] (ADD x x))
+	// cond: c < 31
+	// result: (SLWconst [c+1] x)
+	for {
+		c := auxIntToInt64(v.AuxInt)
+		if v_0.Op != OpPPC64ADD {
+			break
+		}
+		x := v_0.Args[1]
+		if x != v_0.Args[0] || !(c < 31) {
+			break
+		}
+		v.reset(OpPPC64SLWconst)
+		v.AuxInt = int64ToAuxInt(c + 1)
+		v.AddArg(x)
+		return true
+	}
 	// match: (SLWconst [c] z:(MOVWreg x))
 	// cond: c < 32 && buildcfg.GOPPC64 >= 9
 	// result: (EXTSWSLconst [c] x)