code review 81370046: runtime: get rid of most uses of REP for copying/zeroing.

runtime: get rid of most uses of REP for copying/zeroing.

REP MOVSQ and REP STOSQ have a really high startup overhead.
Use a Duff's device to do the repetition instead.

benchmark                 old ns/op     new ns/op     delta
BenchmarkClearFat32       7.20          1.60          -77.78%
BenchmarkCopyFat32        6.88          2.38          -65.41%
BenchmarkClearFat64       7.15          3.20          -55.24%
BenchmarkCopyFat64        6.88          3.44          -50.00%
BenchmarkClearFat128      9.53          5.34          -43.97%
BenchmarkCopyFat128       9.27          5.56          -40.02%
BenchmarkClearFat256      13.8          9.53          -30.94%
BenchmarkCopyFat256       13.5          10.3          -23.70%
BenchmarkClearFat512      22.3          18.0          -19.28%
BenchmarkCopyFat512       22.0          19.7          -10.45%
BenchmarkCopyFat1024      36.5          38.4          +5.21%
BenchmarkClearFat1024     35.1          35.0          -0.28%

TODO: use for stack frame zeroing
TODO: REP prefixes are still used for "reverse" copying when src/dst
regions overlap.  Might be worth fixing.

[khr, 2014-03-27 22:45:44]

Hello golang-codereviews@googlegroups.com,

I'd like you to review this change to
https://khr%40golang.org@code.google.com/p/go/

[khr, 2014-04-01 15:29:57]

On 2014/03/27 22:45:44, khr wrote:
> Hello mailto:golang-codereviews@googlegroups.com,
> 
> I'd like you to review this change to
> https://khr%40golang.org@code.google.com/p/go/

Ping.

[rsc, 2014-04-01 16:50:38]

Sorry, I haven't scanned golang-codereviews in a few days. Please feel free
to add --cc rsc to the 'hg mail' and I will get to them much faster.

[rsc, 2014-04-01 16:55:04]

LGTM

https://codereview.appspot.com/81370046/diff/40001/src/cmd/6g/cgen.c
File src/cmd/6g/cgen.c (right):

https://codereview.appspot.com/81370046/diff/40001/src/cmd/6g/cgen.c#newcode1459
src/cmd/6g/cgen.c:1459: // 14 = magic constant: see
../../pkg/runtime/asm_amd64.s
128 is a magic constant too. :-)

https://codereview.appspot.com/81370046/diff/40001/src/cmd/6l/6.out.h
File src/cmd/6l/6.out.h (right):

https://codereview.appspot.com/81370046/diff/40001/src/cmd/6l/6.out.h#newcode94
src/cmd/6l/6.out.h:94: ADUFFCOPY,
can these go at the end? it doesn't matter much but we've been putting all the
new stuff and pseudo-ops down there.

[khr, 2014-04-01 19:51:04]

*** Submitted as https://code.google.com/p/go/source/detail?r=dd4a24daae43 ***

runtime: get rid of most uses of REP for copying/zeroing.

REP MOVSQ and REP STOSQ have a really high startup overhead.
Use a Duff's device to do the repetition instead.

benchmark                 old ns/op     new ns/op     delta
BenchmarkClearFat32       7.20          1.60          -77.78%
BenchmarkCopyFat32        6.88          2.38          -65.41%
BenchmarkClearFat64       7.15          3.20          -55.24%
BenchmarkCopyFat64        6.88          3.44          -50.00%
BenchmarkClearFat128      9.53          5.34          -43.97%
BenchmarkCopyFat128       9.27          5.56          -40.02%
BenchmarkClearFat256      13.8          9.53          -30.94%
BenchmarkCopyFat256       13.5          10.3          -23.70%
BenchmarkClearFat512      22.3          18.0          -19.28%
BenchmarkCopyFat512       22.0          19.7          -10.45%
BenchmarkCopyFat1024      36.5          38.4          +5.21%
BenchmarkClearFat1024     35.1          35.0          -0.28%

TODO: use for stack frame zeroing
TODO: REP prefixes are still used for "reverse" copying when src/dst
regions overlap.  Might be worth fixing.

LGTM=rsc
R=golang-codereviews, rsc
CC=golang-codereviews, r
https://codereview.appspot.com/81370046

[gobot, 2014-04-01 20:04:03]

This CL appears to have broken the netbsd-amd64-bsiegert builder.
See http://build.golang.org/log/0092e977d88baa4f0a4a7722db7f7a323b981317