cmd/compile: generate Select on WASM

This CL performs the branchelim optimization on WASM with its
select instruction. And the total size of pkg/js_wasm decreased
about 80KB by this optimization.

Change-Id: I868eb146120a1cac5c4609c8e9ddb07e4da8a1d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/190957
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Richard Musiol <neelance@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

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

@@ -20,7 +20,7 @@ package ssa
 func branchelim(f *Func) {
 	// FIXME: add support for lowering CondSelects on more architectures
 	switch f.Config.arch {
-	case "arm64", "amd64":
+	case "arm64", "amd64", "wasm":
 		// implemented
 	default:
 		return

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

@@ -390,6 +390,8 @@
 (PopCount16 x) -> (I64Popcnt (ZeroExt16to64 x))
 (PopCount8  x) -> (I64Popcnt (ZeroExt8to64  x))
 
+(CondSelect <t> x y cond) -> (Select <t> x y cond)
+
 // --- Optimizations ---
 (I64Add (I64Const [x]) (I64Const [y])) -> (I64Const [x + y])
 (I64Mul (I64Const [x]) (I64Const [y])) -> (I64Const [x * y])

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

@@ -50,6 +50,8 @@ func rewriteValueWasm(v *Value) bool {
 		return rewriteValueWasm_OpCom64_0(v)
 	case OpCom8:
 		return rewriteValueWasm_OpCom8_0(v)
+	case OpCondSelect:
+		return rewriteValueWasm_OpCondSelect_0(v)
 	case OpConst16:
 		return rewriteValueWasm_OpConst16_0(v)
 	case OpConst32:
@@ -865,6 +867,23 @@ func rewriteValueWasm_OpCom8_0(v *Value) bool {
 		return true
 	}
 }
+func rewriteValueWasm_OpCondSelect_0(v *Value) bool {
+	// match: (CondSelect <t> x y cond)
+	// cond:
+	// result: (Select <t> x y cond)
+	for {
+		t := v.Type
+		cond := v.Args[2]
+		x := v.Args[0]
+		y := v.Args[1]
+		v.reset(OpWasmSelect)
+		v.Type = t
+		v.AddArg(x)
+		v.AddArg(y)
+		v.AddArg(cond)
+		return true
+	}
+}
 func rewriteValueWasm_OpConst16_0(v *Value) bool {
 	// match: (Const16 [val])
 	// cond:

test/codegen/condmove.go

@@ -13,6 +13,7 @@ func cmovint(c int) int {
 	}
 	// amd64:"CMOVQLT"
 	// arm64:"CSEL\tLT"
+	// wasm:"Select"
 	return x
 }
 
@@ -22,6 +23,7 @@ func cmovchan(x, y chan int) chan int {
 	}
 	// amd64:"CMOVQNE"
 	// arm64:"CSEL\tNE"
+	// wasm:"Select"
 	return x
 }
 
@@ -31,6 +33,7 @@ func cmovuintptr(x, y uintptr) uintptr {
 	}
 	// amd64:"CMOVQCS"
 	// arm64:"CSEL\tLO"
+	// wasm:"Select"
 	return x
 }
 
@@ -40,6 +43,7 @@ func cmov32bit(x, y uint32) uint32 {
 	}
 	// amd64:"CMOVLCS"
 	// arm64:"CSEL\tLO"
+	// wasm:"Select"
 	return x
 }
 
@@ -49,6 +53,7 @@ func cmov16bit(x, y uint16) uint16 {
 	}
 	// amd64:"CMOVWCS"
 	// arm64:"CSEL\tLO"
+	// wasm:"Select"
 	return x
 }
 
@@ -61,6 +66,7 @@ func cmovfloateq(x, y float64) int {
 	}
 	// amd64:"CMOVQNE","CMOVQPC"
 	// arm64:"CSEL\tEQ"
+	// wasm:"Select"
 	return a
 }
 
@@ -71,6 +77,7 @@ func cmovfloatne(x, y float64) int {
 	}
 	// amd64:"CMOVQNE","CMOVQPS"
 	// arm64:"CSEL\tNE"
+	// wasm:"Select"
 	return a
 }
 
@@ -96,6 +103,7 @@ func cmovfloatint2(x, y float64) float64 {
 		}
 		// amd64:"CMOVQHI"
 		// arm64:"CSEL\tMI"
+		// wasm:"Select"
 		r = r - ldexp(y, (rexp-yexp))
 	}
 	return r
@@ -109,6 +117,7 @@ func cmovloaded(x [4]int, y int) int {
 	}
 	// amd64:"CMOVQNE"
 	// arm64:"CSEL\tNE"
+	// wasm:"Select"
 	return y
 }
 
@@ -119,6 +128,7 @@ func cmovuintptr2(x, y uintptr) uintptr {
 	}
 	// amd64:"CMOVQEQ"
 	// arm64:"CSEL\tEQ"
+	// wasm:"Select"
 	return a
 }
 
@@ -130,6 +140,7 @@ func cmovfloatmove(x, y int) float64 {
 	}
 	// amd64:-"CMOV"
 	// arm64:-"CSEL"
+	// wasm:-"Select"
 	return a
 }