Thanks to Minux and Remy for their advice.

The EOR optimisation is applied to a few places in the stdlib.

// hash/crc32/crc32.go
func update(crc uint32, tab *Table, p []byte) uint32 {
	crc = ^crc
	for _, v := range p {
        	crc = tab[byte(crc)^v] ^ (crc >> 8)
	}
	return ^crc
}

before:

--- prog list "update" ---
0164 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:101) TEXT        update+0(SB),$12-24
0165 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:101) MOVW        tab+4(FP),R8
0166 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:102) MOVW        crc+0(FP),R0
0167 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:102) EOR         $-1,R0,R5
0168 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:103) MOVW        p+8(FP),R0
0169 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:103) MOVW        R0,autotmp_0019+-12(SP)

after:

--- prog list "update" ---
0164 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:101) TEXT        update+0(SB),$12-24
0165 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:101) MOVW        tab+4(FP),R8
0166 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:102) MOVW        crc+0(FP),R0
0167 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:102) MVN         R0,R5
0168 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:103) MOVW        p+8(FP),R0
0169 (/home/dfc/go/src/pkg/hash/crc32/crc32.go:103) MOVW        R0,autotmp_0019+-12(SP)

After 5l has done its work,

        crc = ^crc
   3d710:       e59d0014        ldr     r0, [sp, #20]
   3d714:       e3e0b000        mvn     fp, #0
   3d718:       e020500b        eor     r5, r0, fp

becomes

        crc = ^crc
   3d710:       e59d0014        ldr     r0, [sp, #20]
   3d714:       e1e05000        mvn     r5, r0

The MOVB optimisation has a small impact on the stdlib, in strconv
and gzip.

// GZIP (RFC 1952) is little-endian, unlike ZLIB (RFC 1950).
func put2(p []byte, v uint16) {
        p[0] = uint8(v >> 0)
        p[1] = uint8(v >> 8)
}

before:

--- prog list "put2" ---
1369 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:76) TEXT       put2+0(SB),$0-16
1370 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:76) MOVHU      v+12(FP),R4
1371 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVHU      R4,R0
1372 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVHU      R0,R0
1373 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVBU      R0,R1
1374 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVBU      R1,R3
1375 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVW       $p+0(FP),R1

after:

--- prog list "put2" ---
1369 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:76) TEXT       put2+0(SB),$0-16
1370 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:76) MOVHU      v+12(FP),R4
1371 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVHU      R4,R0
1372 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVBU      R0,R1
1373 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVBU      R1,R3
1374 (/home/dfc/go/src/pkg/compress/gzip/gzip.go:77) MOVW       $p+0(FP),R1

R=remyoudompheng, rsc, minux.ma
CC=golang-dev
https://golang.org/cl/6674048