[dev.ssa] cmd/compile/internal/ssa: distribute multiplication into addition

* This is a very basic form of straight line strength reduction.
* Removes one multiplication from a[b].c++; a[b+1].c++
* It increases pressure on the register allocator because
CSE creates more copies of the multiplication sizeof(a[0])*b.

Change-Id: I686a18e9c24cc6f8bdfa925713afed034f7d36d0
Reviewed-on: https://go-review.googlesource.com/20091
Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

2 files changed, 75 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/ssa/gen/generic.rules b/src/cmd/compile/internal/ssa/gen/generic.rules
index ac24337920..11c7b9d7a1 100644
--- a/src/cmd/compile/internal/ssa/gen/generic.rules
+++ b/src/cmd/compile/internal/ssa/gen/generic.rules
@@ -147,6 +147,11 @@
 (Xor16 x (Const16 <t> [c])) && x.Op != OpConst16 -> (Xor16 (Const16 <t> [c]) x)
 (Xor8 x (Const8 <t> [c])) && x.Op != OpConst8 -> (Xor8 (Const8 <t> [c]) x)
 
+// Distribute multiplication c * (d+x) -> c*d + c*x. Useful for:
+// a[i].b = ...; a[i+1].b = ...
+(Mul64 (Const64 <t> [c]) (Add64 <t> (Const64 <t> [d]) x)) -> (Add64 (Const64 <t> [c*d]) (Mul64 <t> (Const64 <t> [c]) x))
+(Mul32 (Const32 <t> [c]) (Add32 <t> (Const32 <t> [d]) x)) -> (Add32 (Const32 <t> [c*d]) (Mul32 <t> (Const32 <t> [c]) x))
+
 // rewrite shifts of 8/16/32 bit consts into 64 bit consts to reduce
 // the number of the other rewrite rules for const shifts
 (Lsh64x32  <t> x (Const32 [c])) -> (Lsh64x64  x (Const64 <t> [int64(uint32(c))]))
diff --git a/src/cmd/compile/internal/ssa/rewritegeneric.go b/src/cmd/compile/internal/ssa/rewritegeneric.go
index 4f29cf5348..0c71b2c884 100644
--- a/src/cmd/compile/internal/ssa/rewritegeneric.go
+++ b/src/cmd/compile/internal/ssa/rewritegeneric.go
@@ -4044,6 +4044,41 @@ func rewriteValuegeneric_OpMul32(v *Value, config *Config) bool {
 		v.AddArg(x)
 		return true
 	}
+	// match: (Mul32 (Const32 <t> [c]) (Add32 <t> (Const32 <t> [d]) x))
+	// cond:
+	// result: (Add32 (Const32 <t> [c*d]) (Mul32 <t> (Const32 <t> [c]) x))
+	for {
+		if v.Args[0].Op != OpConst32 {
+			break
+		}
+		t := v.Args[0].Type
+		c := v.Args[0].AuxInt
+		if v.Args[1].Op != OpAdd32 {
+			break
+		}
+		if v.Args[1].Type != v.Args[0].Type {
+			break
+		}
+		if v.Args[1].Args[0].Op != OpConst32 {
+			break
+		}
+		if v.Args[1].Args[0].Type != v.Args[0].Type {
+			break
+		}
+		d := v.Args[1].Args[0].AuxInt
+		x := v.Args[1].Args[1]
+		v.reset(OpAdd32)
+		v0 := b.NewValue0(v.Line, OpConst32, t)
+		v0.AuxInt = c * d
+		v.AddArg(v0)
+		v1 := b.NewValue0(v.Line, OpMul32, t)
+		v2 := b.NewValue0(v.Line, OpConst32, t)
+		v2.AuxInt = c
+		v1.AddArg(v2)
+		v1.AddArg(x)
+		v.AddArg(v1)
+		return true
+	}
 	// match: (Mul32 (Const32 [0]) _)
 	// cond:
 	// result: (Const32 [0])
@@ -4099,6 +4134,41 @@ func rewriteValuegeneric_OpMul64(v *Value, config *Config) bool {
 		v.AddArg(x)
 		return true
 	}
+	// match: (Mul64 (Const64 <t> [c]) (Add64 <t> (Const64 <t> [d]) x))
+	// cond:
+	// result: (Add64 (Const64 <t> [c*d]) (Mul64 <t> (Const64 <t> [c]) x))
+	for {
+		if v.Args[0].Op != OpConst64 {
+			break
+		}
+		t := v.Args[0].Type
+		c := v.Args[0].AuxInt
+		if v.Args[1].Op != OpAdd64 {
+			break
+		}
+		if v.Args[1].Type != v.Args[0].Type {
+			break
+		}
+		if v.Args[1].Args[0].Op != OpConst64 {
+			break
+		}
+		if v.Args[1].Args[0].Type != v.Args[0].Type {
+			break
+		}
+		d := v.Args[1].Args[0].AuxInt
+		x := v.Args[1].Args[1]
+		v.reset(OpAdd64)
+		v0 := b.NewValue0(v.Line, OpConst64, t)
+		v0.AuxInt = c * d
+		v.AddArg(v0)
+		v1 := b.NewValue0(v.Line, OpMul64, t)
+		v2 := b.NewValue0(v.Line, OpConst64, t)
+		v2.AuxInt = c
+		v1.AddArg(v2)
+		v1.AddArg(x)
+		v.AddArg(v1)
+		return true
+	}
 	// match: (Mul64 (Const64 [0]) _)
 	// cond:
 	// result: (Const64 [0])