[dev.simd] cmd/compile: rewrite to elide Slicemask from len==c>0 slicing

This might have been something that prove could be educated
into figuring out, but this also works, and it also helps
prove downstream.

Adjusted the prove test, because this change moved a message.

Change-Id: I5eabe639eff5db9cd9766a6a8666fdb4973829cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/697715
Commit-Queue: David Chase <drchase@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Bypass: David Chase <drchase@google.com>

2 files changed, 91 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/ssa/_gen/generic.rules b/src/cmd/compile/internal/ssa/_gen/generic.rules
index 00e8fcbe32..8583cfae40 100644
--- a/src/cmd/compile/internal/ssa/_gen/generic.rules
+++ b/src/cmd/compile/internal/ssa/_gen/generic.rules
@@ -989,6 +989,10 @@
     (Const64 <typ.Int> [0])
     (Const64 <typ.Int> [0]))
 
+// Special rule to help constant slicing; len > 0 implies cap > 0 implies Slicemask is all 1
+(SliceMake (AddPtr <t> x (And64 y (Slicemask _))) w:(Const64 [c]) z) && c > 0 => (SliceMake (AddPtr <t> x y) w z)
+(SliceMake (AddPtr <t> x (And32 y (Slicemask _))) w:(Const32 [c]) z) && c > 0 => (SliceMake (AddPtr <t> x y) w z)
+
 // interface ops
 (ConstInterface) =>
   (IMake
diff --git a/src/cmd/compile/internal/ssa/rewritegeneric.go b/src/cmd/compile/internal/ssa/rewritegeneric.go
index fe61ceaff2..c9cff8651b 100644
--- a/src/cmd/compile/internal/ssa/rewritegeneric.go
+++ b/src/cmd/compile/internal/ssa/rewritegeneric.go
@@ -422,6 +422,8 @@ func rewriteValuegeneric(v *Value) bool {
 		return rewriteValuegeneric_OpSliceCap(v)
 	case OpSliceLen:
 		return rewriteValuegeneric_OpSliceLen(v)
+	case OpSliceMake:
+		return rewriteValuegeneric_OpSliceMake(v)
 	case OpSlicePtr:
 		return rewriteValuegeneric_OpSlicePtr(v)
 	case OpSlicemask:
@@ -30645,6 +30647,91 @@ func rewriteValuegeneric_OpSliceLen(v *Value) bool {
 	}
 	return false
 }
+func rewriteValuegeneric_OpSliceMake(v *Value) bool {
+	v_2 := v.Args[2]
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (SliceMake (AddPtr <t> x (And64 y (Slicemask _))) w:(Const64 [c]) z)
+	// cond: c > 0
+	// result: (SliceMake (AddPtr <t> x y) w z)
+	for {
+		if v_0.Op != OpAddPtr {
+			break
+		}
+		t := v_0.Type
+		_ = v_0.Args[1]
+		x := v_0.Args[0]
+		v_0_1 := v_0.Args[1]
+		if v_0_1.Op != OpAnd64 {
+			break
+		}
+		_ = v_0_1.Args[1]
+		v_0_1_0 := v_0_1.Args[0]
+		v_0_1_1 := v_0_1.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_0_1_0, v_0_1_1 = _i0+1, v_0_1_1, v_0_1_0 {
+			y := v_0_1_0
+			if v_0_1_1.Op != OpSlicemask {
+				continue
+			}
+			w := v_1
+			if w.Op != OpConst64 {
+				continue
+			}
+			c := auxIntToInt64(w.AuxInt)
+			z := v_2
+			if !(c > 0) {
+				continue
+			}
+			v.reset(OpSliceMake)
+			v0 := b.NewValue0(v.Pos, OpAddPtr, t)
+			v0.AddArg2(x, y)
+			v.AddArg3(v0, w, z)
+			return true
+		}
+		break
+	}
+	// match: (SliceMake (AddPtr <t> x (And32 y (Slicemask _))) w:(Const32 [c]) z)
+	// cond: c > 0
+	// result: (SliceMake (AddPtr <t> x y) w z)
+	for {
+		if v_0.Op != OpAddPtr {
+			break
+		}
+		t := v_0.Type
+		_ = v_0.Args[1]
+		x := v_0.Args[0]
+		v_0_1 := v_0.Args[1]
+		if v_0_1.Op != OpAnd32 {
+			break
+		}
+		_ = v_0_1.Args[1]
+		v_0_1_0 := v_0_1.Args[0]
+		v_0_1_1 := v_0_1.Args[1]
+		for _i0 := 0; _i0 <= 1; _i0, v_0_1_0, v_0_1_1 = _i0+1, v_0_1_1, v_0_1_0 {
+			y := v_0_1_0
+			if v_0_1_1.Op != OpSlicemask {
+				continue
+			}
+			w := v_1
+			if w.Op != OpConst32 {
+				continue
+			}
+			c := auxIntToInt32(w.AuxInt)
+			z := v_2
+			if !(c > 0) {
+				continue
+			}
+			v.reset(OpSliceMake)
+			v0 := b.NewValue0(v.Pos, OpAddPtr, t)
+			v0.AddArg2(x, y)
+			v.AddArg3(v0, w, z)
+			return true
+		}
+		break
+	}
+	return false
+}
 func rewriteValuegeneric_OpSlicePtr(v *Value) bool {
 	v_0 := v.Args[0]
 	// match: (SlicePtr (SliceMake (SlicePtr x) _ _))