cmd/internal/gc: use MOV R0, R1 instead of LEA 0(R0), R1 in Agen

Minor code generation optimization I've been meaning to do
for a while and noticed while working on the emitted write
barrier code. Using MOV lets the compiler and maybe the
processor do copy propagation.

name                                       old                     new          delta
BenchmarkBinaryTree17              17.9s × (0.99,1.01)     18.0s × (0.99,1.01)  ~
BenchmarkFannkuch11                4.42s × (1.00,1.00)     4.36s × (1.00,1.00)  -1.39%
BenchmarkFmtFprintfEmpty           118ns × (0.96,1.02)     120ns × (0.99,1.06)  ~
BenchmarkFmtFprintfString          486ns × (0.99,1.01)     480ns × (0.99,1.01)  -1.34%
BenchmarkFmtFprintfInt             457ns × (0.99,1.01)     451ns × (0.99,1.01)  -1.31%
BenchmarkFmtFprintfIntInt          768ns × (1.00,1.01)     766ns × (0.99,1.01)  ~
BenchmarkFmtFprintfPrefixedInt     584ns × (0.99,1.03)     569ns × (0.99,1.01)  -2.57%
BenchmarkFmtFprintfFloat           739ns × (0.99,1.00)     728ns × (1.00,1.01)  -1.49%
BenchmarkFmtManyArgs              2.77µs × (1.00,1.00)    2.81µs × (1.00,1.01)  +1.53%
BenchmarkGobDecode                39.3ms × (0.99,1.01)    39.4ms × (0.99,1.01)  ~
BenchmarkGobEncode                39.4ms × (0.99,1.00)    39.4ms × (0.99,1.00)  ~
BenchmarkGzip                      661ms × (0.99,1.01)     660ms × (1.00,1.01)  ~
BenchmarkGunzip                    142ms × (1.00,1.00)     143ms × (1.00,1.00)  +0.20%
BenchmarkHTTPClientServer          133µs × (0.98,1.01)     132µs × (0.99,1.01)  ~
BenchmarkJSONEncode               56.5ms × (0.99,1.01)    57.1ms × (0.99,1.01)  +0.94%
BenchmarkJSONDecode                143ms × (1.00,1.00)     138ms × (1.00,1.01)  -3.22%
BenchmarkMandelbrot200            6.01ms × (1.00,1.00)    6.02ms × (1.00,1.00)  ~
BenchmarkGoParse                  9.63ms × (0.94,1.07)    9.79ms × (0.92,1.07)  ~
BenchmarkRegexpMatchEasy0_32       210ns × (1.00,1.00)     210ns × (1.00,1.01)  ~
BenchmarkRegexpMatchEasy0_1K       596ns × (0.99,1.01)     593ns × (0.99,1.01)  ~
BenchmarkRegexpMatchEasy1_32       184ns × (0.99,1.01)     182ns × (0.99,1.01)  ~
BenchmarkRegexpMatchEasy1_1K      1.01µs × (0.99,1.01)    1.01µs × (1.00,1.01)  ~
BenchmarkRegexpMatchMedium_32      327ns × (1.00,1.01)     331ns × (1.00,1.00)  +1.22%
BenchmarkRegexpMatchMedium_1K     93.0µs × (1.00,1.02)    92.6µs × (1.00,1.01)  ~
BenchmarkRegexpMatchHard_32       4.76µs × (0.95,1.01)    4.58µs × (0.99,1.05)  ~
BenchmarkRegexpMatchHard_1K        136µs × (1.00,1.01)     136µs × (1.00,1.01)  ~
BenchmarkRevcomp                   892ms × (1.00,1.01)     900ms × (0.99,1.06)  ~
BenchmarkTemplate                  175ms × (0.99,1.00)     171ms × (1.00,1.01)  -2.36%
BenchmarkTimeParse                 638ns × (1.00,1.00)     637ns × (1.00,1.00)  ~
BenchmarkTimeFormat                772ns × (1.00,1.00)     742ns × (1.00,1.00)  -3.95%

Change-Id: I6504e310cb9cf48a73d539c478b4dbcacde208b2
Reviewed-on: https://go-review.googlesource.com/9308Reviewed-by: Austin Clements <austin@google.com>

src/cmd/internal/gc/cgen.go

@@ -1512,6 +1512,17 @@ func Agen(n *Node, res *Node) {
 		return
 	}
 
+	if n.Op == OINDREG && n.Xoffset == 0 {
+		// Generate MOVW R0, R1 instead of MOVW $0(R0), R1.
+		// This allows better move propagation in the back ends
+		// (and maybe it helps the processor).
+		n1 := *n
+		n1.Op = OREGISTER
+		n1.Type = res.Type
+		Thearch.Gmove(&n1, res)
+		return
+	}
+
 	if n.Addable {
 		if n.Op == OREGISTER {
 			Fatal("agen OREGISTER")