cmd/compile: strength reduce *24

We use *24 a lot for pointer arithmetic when accessing slices
of slices ([][]T).  Rewrite to use an LEA and a shift.
The shift will likely be free, as it often gets folded into
an indexed load/store.

Update #14606

Change-Id: Ie0bf6dc1093876efd57e88ce5f62c26a9bf21cec
Reviewed-on: https://go-review.googlesource.com/20567
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Todd Neal <todd@tneal.org>

src/cmd/compile/internal/ssa/gen/AMD64.rules

@@ -548,6 +548,7 @@
 (MULQconst [3] x) -> (LEAQ2 x x)
 (MULQconst [5] x) -> (LEAQ4 x x)
 (MULQconst [9] x) -> (LEAQ8 x x)
+(MULQconst [24] x) -> (SHLQconst [3] (LEAQ2 <v.Type> x x)) // Useful for [][]T accesses
 (MULQconst [c] x) && isPowerOfTwo(c) -> (SHLQconst [log2(c)] x)
 
 // combine add/shift into LEAQ

src/cmd/compile/internal/ssa/rewriteAMD64.go

@@ -8280,6 +8280,22 @@ func rewriteValueAMD64_OpAMD64MULQconst(v *Value, config *Config) bool {
 		v.AddArg(x)
 		return true
 	}
+	// match: (MULQconst [24] x)
+	// cond:
+	// result: (SHLQconst [3] (LEAQ2 <v.Type> x x))
+	for {
+		if v.AuxInt != 24 {
+			break
+		}
+		x := v.Args[0]
+		v.reset(OpAMD64SHLQconst)
+		v.AuxInt = 3
+		v0 := b.NewValue0(v.Line, OpAMD64LEAQ2, v.Type)
+		v0.AddArg(x)
+		v0.AddArg(x)
+		v.AddArg(v0)
+		return true
+	}
 	// match: (MULQconst [c] x)
 	// cond: isPowerOfTwo(c)
 	// result: (SHLQconst [log2(c)] x)