cmd/5g, cmd/6g, cmd/8g: clear Addr node when registerizing

cmd/5g, cmd/6g, cmd/8g: clear Addr node when registerizing

Update issue 8525

Some temporary variables that were fully registerized nevertheless had stack space allocated for them because the Addrs were still marked as having associated nodes.

Distribution of stack space reserved for temporary variables while running make.bash (6g):

BEFORE

40.89%  7026 allocauto: 0 to 0
 7.83%  1346 allocauto: 0 to 24
 7.22%  1241 allocauto: 0 to 8
 6.30%  1082 allocauto: 0 to 16
 4.96%   853 allocauto: 0 to 56
 4.59%   789 allocauto: 0 to 32
 2.97%   510 allocauto: 0 to 40
 2.32%   399 allocauto: 0 to 48
 2.10%   360 allocauto: 0 to 64
 1.91%   328 allocauto: 0 to 72

AFTER

48.49%  8332 allocauto: 0 to 0
 9.52%  1635 allocauto: 0 to 16
 5.28%   908 allocauto: 0 to 48
 4.80%   824 allocauto: 0 to 32
 4.73%   812 allocauto: 0 to 8
 3.38%   581 allocauto: 0 to 24
 2.35%   404 allocauto: 0 to 40
 2.32%   399 allocauto: 0 to 64
 1.65%   284 allocauto: 0 to 56
 1.34%   230 allocauto: 0 to 72

[josharian, 2014-08-24 18:51:09]

Hello rsc@golang.org (cc: dave@cheney.net, dvyukov@google.com,
golang-codereviews@googlegroups.com, minux@golang.org),

I'd like you to review this change to
https://code.google.com/p/go

[josharian, 2014-08-24 18:52:14]

cc minux for 9g

dfc, will you test on arm for me? I don't have my hardware available at the
moment. Thanks.

[josharian, 2014-08-24 19:30:49]

Sigh. Scratch 5g for now. The corresponding change causes

pkg/runtime/debug.go:69: internal error: (*MemProfileRecord).Stack ~r0 (type
[]uintptr) recorded as live on entry

and many more, and I don't understand why.

[josharian, 2014-08-24 23:37:22]

> Sigh. Scratch 5g for now. The corresponding change causes

Looks like 5g (unlike 6g and 8g) handle temp cleanup at the end of regopt, after
the peephole pass. So this is ready for review -- 6g and 8g only, on purpose.

[rsc, 2014-08-25 01:05:10]

Thanks for finding this.

The fix is papering over the real problem, namely that there are Addr structs
with a register type that have node set anyway (it should be cleared). I believe
the real fix is to edit addreg to set a->node = nil after it sets a->sym = nil.
I also believe this should work fine on all architectures.

[josharian, 2014-08-25 01:57:29]

PTAL

Thanks as always, rsc, for all your help.

[rsc, 2014-08-25 02:04:45]

LGTM

[josharian, 2014-08-25 02:07:33]

*** Submitted as https://code.google.com/p/go/source/detail?r=2cd1e4d7c117 ***

cmd/5g, cmd/6g, cmd/8g: clear Addr node when registerizing

Update issue 8525

Some temporary variables that were fully registerized nevertheless had stack
space allocated for them because the Addrs were still marked as having
associated nodes.

Distribution of stack space reserved for temporary variables while running
make.bash (6g):

BEFORE

40.89%  7026 allocauto: 0 to 0
 7.83%  1346 allocauto: 0 to 24
 7.22%  1241 allocauto: 0 to 8
 6.30%  1082 allocauto: 0 to 16
 4.96%   853 allocauto: 0 to 56
 4.59%   789 allocauto: 0 to 32
 2.97%   510 allocauto: 0 to 40
 2.32%   399 allocauto: 0 to 48
 2.10%   360 allocauto: 0 to 64
 1.91%   328 allocauto: 0 to 72

AFTER

48.49%  8332 allocauto: 0 to 0
 9.52%  1635 allocauto: 0 to 16
 5.28%   908 allocauto: 0 to 48
 4.80%   824 allocauto: 0 to 32
 4.73%   812 allocauto: 0 to 8
 3.38%   581 allocauto: 0 to 24
 2.35%   404 allocauto: 0 to 40
 2.32%   399 allocauto: 0 to 64
 1.65%   284 allocauto: 0 to 56
 1.34%   230 allocauto: 0 to 72

LGTM=rsc
R=rsc
CC=dave, dvyukov, golang-codereviews, minux
https://codereview.appspot.com/126160043