cmd/compile: fold negation into multiplication on loong64

This change also add corresponding benchmark tests and codegen tests.
The performance improvement on CPU Loongson-3A6000-HV is as follows:

goos: linux
goarch: loong64
pkg: cmd/compile/internal/test
cpu: Loongson-3A6000-HV @ 2500.00MHz
        |  bench.old   |              bench.new              |
        |    sec/op    |   sec/op     vs base                |
MulNeg     828.4n ± 0%   655.9n ± 0%  -20.82% (p=0.000 n=10)
Mul2Neg   1062.0n ± 0%   826.8n ± 0%  -22.15% (p=0.000 n=10)
geomean    938.0n        736.4n       -21.49%

Change-Id: Ia999732880ec65be0c66cddc757a4868847e5b15
Reviewed-on: https://go-review.googlesource.com/c/go/+/682535
Reviewed-by: Dmitri Shuralyov <[email protected]>
Reviewed-by: abner chenc <[email protected]>
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Mark Freeman <[email protected]>

Author: sophie-zhao

src/cmd/compile/internal/ssa/_gen/LOONG64.rules

@@ -755,6 +755,9 @@
 
 (MULV  x (MOVVconst [c])) && canMulStrengthReduce(config, c) => {mulStrengthReduce(v, x, c)}
 
+(MULV (NEGV x) (MOVVconst [c])) => (MULV x (MOVVconst [-c]))
+(MULV (NEGV x) (NEGV y)) => (MULV x y)
+
 // div by constant
 (DIVVU x (MOVVconst [1])) => x
 (DIVVU x (MOVVconst [c])) && isPowerOfTwo(c) => (SRLVconst [log64(c)] x)

src/cmd/compile/internal/ssa/rewriteLOONG64.go

@@ -122,3 +122,26 @@ func BenchmarkBitToggleConst(b *testing.B) {
 		}
 	}
 }
+
+func BenchmarkMulNeg(b *testing.B) {
+	x := make([]int64, 1024)
+	for i := 0; i < b.N; i++ {
+		var s int64
+		for i := range x {
+			s = (-x[i]) * 11
+		}
+		globl = s
+	}
+}
+
+func BenchmarkMul2Neg(b *testing.B) {
+	x := make([]int64, 1024)
+	y := make([]int64, 1024)
+	for i := 0; i < b.N; i++ {
+		var s int64
+		for i := range x {
+			s = (-x[i]) * (-y[i])
+		}
+		globl = s
+	}
+}

src/cmd/compile/internal/test/bench_test.go

@@ -314,6 +314,18 @@ func MergeMuls5(a, n int) int {
 	return a*n - 19*n // (a-19)n
 }
 
+// Multiplications folded negation
+
+func FoldNegMul(a int) int {
+	// loong64:"MULV","MOVV\t[$]-11",-"SUBVU\tR[0-9], R0,"
+	return (-a) * 11
+}
+
+func Fold2NegMul(a, b int) int {
+	// loong64:"MULV",-"SUBVU\tR[0-9], R0,"
+	return (-a) * (-b)
+}
+
 // -------------- //
 //    Division    //
 // -------------- //