cmd/compile: set unalignedOK to make memcombine work properly on loong64

goos: linux
goarch: loong64
pkg: unicode/utf8
cpu: Loongson-3A6000-HV @ 2500.00MHz
                            │     old     │                 new                 │
                            │   sec/op    │   sec/op     vs base                │
ValidTenASCIIChars            7.604n ± 0%   6.805n ± 0%  -10.51% (p=0.000 n=10)
Valid100KASCIIChars           37.41µ ± 0%   16.58µ ± 0%  -55.67% (p=0.000 n=10)
ValidTenJapaneseChars         60.84n ± 0%   58.62n ± 0%   -3.64% (p=0.000 n=10)
ValidLongMostlyASCII          113.5µ ± 0%   113.5µ ± 0%        ~ (p=0.303 n=10)
ValidLongJapanese             204.6µ ± 0%   206.8µ ± 0%   +1.07% (p=0.000 n=10)
ValidStringTenASCIIChars      7.604n ± 0%   6.803n ± 0%  -10.53% (p=0.000 n=10)
ValidString100KASCIIChars     38.05µ ± 0%   17.14µ ± 0%  -54.97% (p=0.000 n=10)
ValidStringTenJapaneseChars   60.58n ± 0%   59.48n ± 0%   -1.82% (p=0.000 n=10)
ValidStringLongMostlyASCII    113.5µ ± 0%   113.4µ ± 0%   -0.10% (p=0.000 n=10)
ValidStringLongJapanese       205.9µ ± 0%   207.3µ ± 0%   +0.67% (p=0.000 n=10)
geomean                       3.324µ        2.756µ       -17.08%

Change-Id: Id43b6e2e41907bd4b92f421dacde31f048db47d6
Reviewed-on: https://go-review.googlesource.com/c/go/+/662495
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Dmitri Shuralyov <[email protected]>
Reviewed-by: Keith Randall <[email protected]>
Auto-Submit: Keith Randall <[email protected]>
Reviewed-by: abner chenc <[email protected]>
Reviewed-by: Keith Randall <[email protected]>

Author: limeidan

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

@@ -283,6 +283,7 @@ func NewConfig(arch string, types Types, ctxt *obj.Link, optimize, softfloat boo
 		c.FPReg = framepointerRegLOONG64
 		c.LinkReg = linkRegLOONG64
 		c.hasGReg = true
+		c.unalignedOK = true
 	case "s390x":
 		c.PtrSize = 8
 		c.RegSize = 8

src/cmd/internal/sys/arch.go

@@ -144,7 +144,7 @@ var ArchLoong64 = &Arch{
 	RegSize:        8,
 	MinLC:          4,
 	Alignment:      8, // Unaligned accesses are not guaranteed to be fast
-	CanMergeLoads:  false,
+	CanMergeLoads:  true,
 	HasLR:          true,
 	FixedFrameSize: 8, // LR
 }

test/codegen/memcombine.go

@@ -19,7 +19,7 @@ func load_le64(b []byte) uint64 {
 	// amd64:`MOVQ\s\(.*\),`,-`MOV[BWL]\t[^$]`,-`OR`
 	// s390x:`MOVDBR\s\(.*\),`
 	// arm64:`MOVD\s\(R[0-9]+\),`,-`MOV[BHW]`
-	// loong64:`MOVBU\s\(R[0-9]+\),`
+	// loong64:`MOVV\s\(R[0-9]+\),`
 	// ppc64le:`MOVD\s`,-`MOV[BHW]Z`
 	// ppc64:`MOVDBR\s`,-`MOV[BHW]Z`
 	return binary.LittleEndian.Uint64(b)
@@ -29,7 +29,7 @@ func load_le64_idx(b []byte, idx int) uint64 {
 	// amd64:`MOVQ\s\(.*\)\(.*\*1\),`,-`MOV[BWL]\t[^$]`,-`OR`
 	// s390x:`MOVDBR\s\(.*\)\(.*\*1\),`
 	// arm64:`MOVD\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[BHW]`
-	// loong64:`MOVBU\s\(R[0-9]+\)\(R[0-9]+\),`
+	// loong64:`MOVV\s\(R[0-9]+\)\(R[0-9]+\),`
 	// ppc64le:`MOVD\s`,-`MOV[BHW]Z\s`
 	// ppc64:`MOVDBR\s`,-`MOV[BHW]Z\s`
 	return binary.LittleEndian.Uint64(b[idx:])
@@ -40,7 +40,7 @@ func load_le32(b []byte) uint32 {
 	// 386:`MOVL\s\(.*\),`,-`MOV[BW]`,-`OR`
 	// s390x:`MOVWBR\s\(.*\),`
 	// arm64:`MOVWU\s\(R[0-9]+\),`,-`MOV[BH]`
-	// loong64:`MOVBU\s\(R[0-9]+\),`
+	// loong64:`MOVWU\s\(R[0-9]+\),`
 	// ppc64le:`MOVWZ\s`,-`MOV[BH]Z\s`
 	// ppc64:`MOVWBR\s`,-`MOV[BH]Z\s`
 	return binary.LittleEndian.Uint32(b)
@@ -51,7 +51,7 @@ func load_le32_idx(b []byte, idx int) uint32 {
 	// 386:`MOVL\s\(.*\)\(.*\*1\),`,-`MOV[BW]`,-`OR`
 	// s390x:`MOVWBR\s\(.*\)\(.*\*1\),`
 	// arm64:`MOVWU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[BH]`
-	// loong64:`MOVBU\s\(R[0-9]+\)\(R[0-9]+\),`
+	// loong64:`MOVWU\s\(R[0-9]+\)\(R[0-9]+\),`
 	// ppc64le:`MOVWZ\s`,-`MOV[BH]Z\s`
 	// ppc64:`MOVWBR\s`,-`MOV[BH]Z\s'
 	return binary.LittleEndian.Uint32(b[idx:])
@@ -61,7 +61,7 @@ func load_le16(b []byte) uint16 {
 	// amd64:`MOVWLZX\s\(.*\),`,-`MOVB`,-`OR`
 	// ppc64le:`MOVHZ\s`,-`MOVBZ`
 	// arm64:`MOVHU\s\(R[0-9]+\),`,-`MOVB`
-	// loong64:`MOVBU\s\(R[0-9]+\),`
+	// loong64:`MOVHU\s\(R[0-9]+\),`
 	// s390x:`MOVHBR\s\(.*\),`
 	// ppc64:`MOVHBR\s`,-`MOVBZ`
 	return binary.LittleEndian.Uint16(b)
@@ -72,7 +72,7 @@ func load_le16_idx(b []byte, idx int) uint16 {
 	// ppc64le:`MOVHZ\s`,-`MOVBZ`
 	// ppc64:`MOVHBR\s`,-`MOVBZ`
 	// arm64:`MOVHU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOVB`
-	// loong64:`MOVBU\s\(R[0-9]+\)\(R[0-9]+\),`
+	// loong64:`MOVHU\s\(R[0-9]+\)\(R[0-9]+\),`
 	// s390x:`MOVHBR\s\(.*\)\(.*\*1\),`
 	return binary.LittleEndian.Uint16(b[idx:])
 }