What version of Go are you using (go version)?

1.9.2

Does this issue reproduce with the latest release?

yes

What operating system and processor architecture are you using (go env)?

darwin_amd64

What did you do?

Optimization for map[i] op= r, except /= and %= (it has possibility of /= 0 and %= 0 that we can not optimize using proposed strategy)

What did you expect to see?

"map[k] op= r" should be using only single call:
func mapassign(t *maptype, h *hmap, key unsafe.Pointer) unsafe.Pointer

What did you see instead?

"map[k] op= r" is using:
func mapaccess*(..) unsafe.Pointer
func mapassign*(...) unsafe.Pointer

We can avoid mapaccess in most cases (except /=).

Description (will simplify some details that are not important here):

This is slightly related to #21687

Currently orderexpr performs transformation of node with OP OASOP (op=) into OAS (=):

l op= r into l = l op r

For map
m[k] op= r

It creates safe expressions:

tmp1 := m[k]
m[k] = tmp1 op r

then walkexpr replaces it with:

tmp1 := mapaccess(m, k) // this never changes m but returns default value if necessary.
tmp2 := massign(m, k) // inserts if necessary
tmp2 = tmp1 op r

We always evaluate r before massign (see test fixedbugs/issue22881.go).

Consider this, mapaccess is not necessary.

Except the m[k] /= r , you may see example at fixedbugs/issue22881.go, f6 where r == 0 causes panic. Other Op doesn't have this problem.

Implementation trick is to do it during walkexpr when m[k] is already replaced with massign.