test/codegen: add arithmetic tests for 386/amd64/arm/arm64

This CL adds several test cases of arithmetic operations for
386/amd64/arm/arm64.

Change-Id: I362687c06249f31091458a1d8c45fc4d006b616a
Reviewed-on: https://go-review.googlesource.com/c/151897
Run-TryBot: Ben Shi <powerman1st@163.com>
Reviewed-by: Keith Randall <khr@golang.org>

test/codegen/arithmetic.go

@@ -66,7 +66,10 @@ func Pow2Muls(n1, n2 int) (int, int) {
 }
 
 func Mul_96(n int) int {
-	// amd64:`SHLQ\t[$]5`,`LEAQ\t\(.*\)\(.*\*2\),`
+	// amd64:`SHLQ\t[$]5`,`LEAQ\t\(.*\)\(.*\*2\),`,-`IMULQ`
+	// 386:`SHLL\t[$]5`,`LEAL\t\(.*\)\(.*\*2\),`,-`IMULL`
+	// arm64:`LSL\t[$]5`,`ADD\sR[0-9]+<<1,\sR[0-9]+`,-`MUL`
+	// arm:`SLL\t[$]5`,`ADD\sR[0-9]+<<1,\sR[0-9]+`,-`MUL`
 	return n * 96
 }
 
@@ -74,6 +77,7 @@ func MulMemSrc(a []uint32, b []float32) {
 	// 386:`IMULL\s4\([A-Z]+\),\s[A-Z]+`
 	a[0] *= a[1]
 	// 386/sse2:`MULSS\s4\([A-Z]+\),\sX[0-9]+`
+	// amd64:`MULSS\s4\([A-Z]+\),\sX[0-9]+`
 	b[0] *= b[1]
 }
 
@@ -115,6 +119,7 @@ func MergeMuls5(a, n int) int {
 
 func DivMemSrc(a []float64) {
 	// 386/sse2:`DIVSD\s8\([A-Z]+\),\sX[0-9]+`
+	// amd64:`DIVSD\s8\([A-Z]+\),\sX[0-9]+`
 	a[0] /= a[1]
 }
 
@@ -141,9 +146,15 @@ func Pow2Divs(n1 uint, n2 int) (uint, int) {
 // Check that constant divisions get turned into MULs
 func ConstDivs(n1 uint, n2 int) (uint, int) {
 	// amd64:"MOVQ\t[$]-1085102592571150095","MULQ",-"DIVQ"
+	// 386:"MOVL\t[$]-252645135","MULL",-"DIVL"
+	// arm64:`MOVD`,`UMULH`,-`DIV`
+	// arm:`MOVW`,`MUL`,-`.*udiv`
 	a := n1 / 17 // unsigned
 
 	// amd64:"MOVQ\t[$]-1085102592571150095","IMULQ",-"IDIVQ"
+	// 386:"MOVL\t[$]-252645135","IMULL",-"IDIVL"
+	// arm64:`MOVD`,`SMULH`,-`DIV`
+	// arm:`MOVW`,`MUL`,-`.*udiv`
 	b := n2 / 17 // signed
 
 	return a, b
@@ -151,6 +162,7 @@ func ConstDivs(n1 uint, n2 int) (uint, int) {
 
 func FloatDivs(a []float32) float32 {
 	// amd64:`DIVSS\s8\([A-Z]+\),\sX[0-9]+`
+	// 386/sse2:`DIVSS\s8\([A-Z]+\),\sX[0-9]+`
 	return a[1] / a[2]
 }
 
@@ -175,9 +187,15 @@ func Pow2Mods(n1 uint, n2 int) (uint, int) {
 // Check that constant modulo divs get turned into MULs
 func ConstMods(n1 uint, n2 int) (uint, int) {
 	// amd64:"MOVQ\t[$]-1085102592571150095","MULQ",-"DIVQ"
+	// 386:"MOVL\t[$]-252645135","MULL",-"DIVL"
+	// arm64:`MOVD`,`UMULH`,-`DIV`
+	// arm:`MOVW`,`MUL`,-`.*udiv`
 	a := n1 % 17 // unsigned
 
 	// amd64:"MOVQ\t[$]-1085102592571150095","IMULQ",-"IDIVQ"
+	// 386:"MOVL\t[$]-252645135","IMULL",-"IDIVL"
+	// arm64:`MOVD`,`SMULH`,-`DIV`
+	// arm:`MOVW`,`MUL`,-`.*udiv`
 	b := n2 % 17 // signed
 
 	return a, b
@@ -270,6 +288,8 @@ func NoFix16B(divd int16) (int16, int16) {
 func LenDiv1(a []int) int {
 	// 386:"SHRL\t[$]10"
 	// amd64:"SHRQ\t[$]10"
+	// arm64:"LSR\t[$]10",-"SDIV"
+	// arm:"SRL\t[$]10",-".*udiv"
 	// ppc64:"SRD"\t[$]10"
 	// ppc64le:"SRD"\t[$]10"
 	return len(a) / 1024
@@ -278,6 +298,8 @@ func LenDiv1(a []int) int {
 func LenDiv2(s string) int {
 	// 386:"SHRL\t[$]11"
 	// amd64:"SHRQ\t[$]11"
+	// arm64:"LSR\t[$]11",-"SDIV"
+	// arm:"SRL\t[$]11",-".*udiv"
 	// ppc64:"SRD\t[$]11"
 	// ppc64le:"SRD\t[$]11"
 	return len(s) / (4097 >> 1)
@@ -286,6 +308,9 @@ func LenDiv2(s string) int {
 func LenMod1(a []int) int {
 	// 386:"ANDL\t[$]1023"
 	// amd64:"ANDQ\t[$]1023"
+	// arm64:"AND\t[$]1023",-"SDIV"
+	// arm/6:"AND",-".*udiv"
+	// arm/7:"BFC",-".*udiv",-"AND"
 	// ppc64:"ANDCC\t[$]1023"
 	// ppc64le:"ANDCC\t[$]1023"
 	return len(a) % 1024
@@ -294,6 +319,9 @@ func LenMod1(a []int) int {
 func LenMod2(s string) int {
 	// 386:"ANDL\t[$]2047"
 	// amd64:"ANDQ\t[$]2047"
+	// arm64:"AND\t[$]2047",-"SDIV"
+	// arm/6:"AND",-".*udiv"
+	// arm/7:"BFC",-".*udiv",-"AND"
 	// ppc64:"ANDCC\t[$]2047"
 	// ppc64le:"ANDCC\t[$]2047"
 	return len(s) % (4097 >> 1)
@@ -302,6 +330,8 @@ func LenMod2(s string) int {
 func CapDiv(a []int) int {
 	// 386:"SHRL\t[$]12"
 	// amd64:"SHRQ\t[$]12"
+	// arm64:"LSR\t[$]12",-"SDIV"
+	// arm:"SRL\t[$]12",-".*udiv"
 	// ppc64:"SRD\t[$]12"
 	// ppc64le:"SRD\t[$]12"
 	return cap(a) / ((1 << 11) + 2048)
@@ -310,6 +340,9 @@ func CapDiv(a []int) int {
 func CapMod(a []int) int {
 	// 386:"ANDL\t[$]4095"
 	// amd64:"ANDQ\t[$]4095"
+	// arm64:"AND\t[$]4095",-"SDIV"
+	// arm/6:"AND",-".*udiv"
+	// arm/7:"BFC",-".*udiv",-"AND"
 	// ppc64:"ANDCC\t[$]4095"
 	// ppc64le:"ANDCC\t[$]4095"
 	return cap(a) % ((1 << 11) + 2048)