]> Cypherpunks repositories - gostls13.git/log
gostls13.git
9 years agocmd/go: fix "#cgo pkg-config:" comments with gccgo
Michael Hudson-Doyle [Thu, 21 Jan 2016 01:53:55 +0000 (14:53 +1300)]
cmd/go: fix "#cgo pkg-config:" comments with gccgo

The unique difficulty of #cgo pkg-config is that the linker flags are recorded
when the package is compiled but (obviously) must be used when the package is
linked into an executable -- so the flags need to be stored on disk somewhere.
As it happens cgo already writes out a _cgo_flags file: nothing uses it
currently, but this change adds it to the lib$pkg.a file when compiling a
package, reads it out when linking (and passes a version of the .a file with
_cgo_flags stripped out of it to the linker). It's all fairly ugly but it works
and I can't really think of any way of reducing the essential level of
ugliness.

Fixes #11739

Change-Id: I35621878014e1e107eda77a5b0b23d0240ec5750
Reviewed-on: https://go-review.googlesource.com/18790
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocrypto/aes: fix vet warnings in gcm_amd64.s
Josh Bleecher Snyder [Thu, 14 Apr 2016 19:41:54 +0000 (12:41 -0700)]
crypto/aes: fix vet warnings in gcm_amd64.s

Notably, this fixes two incorrect argument sizes.

Update #11041

Change-Id: Ie4a3b1a59cd6a6707f6d2f4d3be978fc70322b46
Reviewed-on: https://go-review.googlesource.com/22091
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agodoc: GCC 6 will have the Go 1.6.1 user libraries
Ian Lance Taylor [Thu, 14 Apr 2016 04:44:35 +0000 (21:44 -0700)]
doc: GCC 6 will have the Go 1.6.1 user libraries

Update #14759.

Change-Id: I8a174aad721beb62380e10071d9648b6b1c21b8c
Reviewed-on: https://go-review.googlesource.com/22072
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/compile: use type. prefix on importpath symbol
David Crawshaw [Thu, 14 Apr 2016 12:54:15 +0000 (08:54 -0400)]
cmd/compile: use type. prefix on importpath symbol

This ensures that importpath symbols are treated like other type data
and end up in the same section under all build modes.

Fixes: go test -buildmode=pie reflect
Change-Id: Ibb8348648e8dcc850f2424d206990a06090ce4c6
Reviewed-on: https://go-review.googlesource.com/22081
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile: clear hidden value at end of channel range body
Josh Bleecher Snyder [Thu, 14 Apr 2016 15:48:36 +0000 (08:48 -0700)]
cmd/compile: clear hidden value at end of channel range body

While we’re here, clean up a few comments.

Fixes #15281

Change-Id: Ia6173e9941133db08f57bc80bdd3c5722122bfdb
Reviewed-on: https://go-review.googlesource.com/22082
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
9 years agotest: use correct value in error message in init1.go
Michael Munday [Thu, 14 Apr 2016 18:07:59 +0000 (14:07 -0400)]
test: use correct value in error message in init1.go

Print numGC followed by numGC1, rather than printing numGC twice.

Change-Id: I8e7144b6a11d4ae9be0d82d88b86fed04b906e2f
Reviewed-on: https://go-review.googlesource.com/22087
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoruntime: find a home for orphaned comments
Jeremy Jackins [Wed, 13 Apr 2016 09:16:21 +0000 (18:16 +0900)]
runtime: find a home for orphaned comments

These comments were left behind after runtime.h was converted
from C to Go. I examined the original code and tried to move these
to the places that the most sense.

Change-Id: I8769d60234c0113d682f9de3bd8d6c34c450c188
Reviewed-on: https://go-review.googlesource.com/21969
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocrypto/cipher: enable fastXORBytes on s390x
Michael Munday [Thu, 14 Apr 2016 17:29:37 +0000 (13:29 -0400)]
crypto/cipher: enable fastXORBytes on s390x

s390x can handle unaligned loads and stores of 64-bit values.

Change-Id: Iae5621781e3ba56e27b4a1f4788772c86e4f6475
Reviewed-on: https://go-review.googlesource.com/22086
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: fix use of original spill name after sinking
David Chase [Wed, 13 Apr 2016 17:30:03 +0000 (13:30 -0400)]
cmd/compile: fix use of original spill name after sinking

This is a fix for the ssacheck builder
http://build.golang.org/log/baa00f70c34e41186051cfe90568de3d91f115d7
after CL 21307 for sinking spills down loop exits
https://go-review.googlesource.com/#/c/21037/

The fix is to reuse (move) the original spill, thus preserving
the definition of the variable and its use count. Original and
copy both use the same stack slot, but ssacheck needs to see
a definition for the variable itself.

Fixes #15279.

Change-Id: I286285490193dc211b312d64dbc5a54867730bd6
Reviewed-on: https://go-review.googlesource.com/21995
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agogo/types: record CallExpr result type even if argument is invalid
Alan Donovan [Thu, 14 Apr 2016 17:41:32 +0000 (13:41 -0400)]
go/types: record CallExpr result type even if argument is invalid

+ test

Fixes #15305

Change-Id: Ica657c00c92f0b19f0df7452cdbe5a95d23cc8a4
Reviewed-on: https://go-review.googlesource.com/22085
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agocmd/vet: teach asm checker about PEXTRD’s op size
Josh Bleecher Snyder [Thu, 14 Apr 2016 16:39:40 +0000 (09:39 -0700)]
cmd/vet: teach asm checker about PEXTRD’s op size

Fixes #15271

Change-Id: I28e3fb5bde1e6fd5b263b1434873b8ce051aee97
Reviewed-on: https://go-review.googlesource.com/22083
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
9 years agocmd/internal/obj: remove use of package bio
Matthew Dempsky [Wed, 13 Apr 2016 00:58:46 +0000 (17:58 -0700)]
cmd/internal/obj: remove use of package bio

Also add MustClose and MustWriter to cmd/internal/bio, and use them in
cmd/asm.

Change-Id: I07f5df3b66c17bc5b2e6ec9c4357d9b653e354e0
Reviewed-on: https://go-review.googlesource.com/21938
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocrypto/x509: don't add an AuthorityKeyId to self-signed certificates.
Adam Langley [Tue, 12 Apr 2016 18:14:25 +0000 (11:14 -0700)]
crypto/x509: don't add an AuthorityKeyId to self-signed certificates.

The AuthorityKeyId is optional for self-signed certificates, generally
useless, and takes up space. This change causes an AuthorityKeyId not to
be added to self-signed certificates, although it can still be set in
the template if the caller really wants to include it.

Fixes #15194.

Change-Id: If5d3c3d9ca9ae5fe67458291510ec7140829756e
Reviewed-on: https://go-review.googlesource.com/21895
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocrypto/tls: make error prefix uniform.
Adam Langley [Tue, 12 Apr 2016 17:43:44 +0000 (10:43 -0700)]
crypto/tls: make error prefix uniform.

Error strings in this package were all over the place: some were
prefixed with “tls:”, some with “crypto/tls:” and some didn't have a
prefix.

This change makes everything use the prefix “tls:”.

Change-Id: Ie8b073c897764b691140412ecd6613da8c4e33a2
Reviewed-on: https://go-review.googlesource.com/21893
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

9 years agocmd/vet: allow untyped composite literals to be unkeyed
Luan Santos [Wed, 13 Apr 2016 15:10:41 +0000 (08:10 -0700)]
cmd/vet: allow untyped composite literals to be unkeyed

We can trust that untyped composite literals are part of a slice literal
and not emit a vet warning for those.

Fixes #9171

Change-Id: Ia7c081e543b850f8be1fd1f9e711520061e70bed
Reviewed-on: https://go-review.googlesource.com/22000
Run-TryBot: Rob Pike <r@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
9 years agomisc/trace: update trace viewer html
Dmitry Vyukov [Thu, 14 Apr 2016 14:31:42 +0000 (16:31 +0200)]
misc/trace: update trace viewer html

The old trace-viewer is broken since Chrome 49:
https://bugs.chromium.org/p/chromium/issues/detail?id=569417
It was fixed in:
https://github.com/catapult-project/catapult/commit/506457cbd726324f327b80ae11f46c1dfeb8710d

This change updates trace-viewer to the latest version
(now it is called catapult).

This version has a bug in the lean config that we use, though:
https://github.com/catapult-project/catapult/issues/2247
So use full config for now (it works, but leads to larger html).
When the bug is fixed we need to switch back to lean config (issue #15302).

Change-Id: Ifb8d782ced66e3292d81c5604039fe18eaf267c5
Reviewed-on: https://go-review.googlesource.com/22013
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agotesting: removed flakey test
Marcel van Lohuizen [Thu, 14 Apr 2016 07:44:48 +0000 (15:44 +0800)]
testing: removed flakey test

The synchronization in this test is a bit complicated and likely
incorrect, judging from the sporadically hanging trybots.
Most of what this is supposed to test is already tested in
TestTestContext, so I'll just remove it.

Fixes #15170

Change-Id: If54db977503caa109cec4516974eda9191051888
Reviewed-on: https://go-review.googlesource.com/22080
Run-TryBot: Marcel van Lohuizen <mpvl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet: fix TestDialAddrError
Mikio Hara [Thu, 14 Apr 2016 05:53:12 +0000 (14:53 +0900)]
net: fix TestDialAddrError

Fixes #15291.

Change-Id: I563140c2acd37d4989a940488b217414cf73f6c2
Reviewed-on: https://go-review.googlesource.com/22077
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
9 years agonet: make use of internal/testenv package
Mikio Hara [Thu, 14 Apr 2016 03:17:44 +0000 (12:17 +0900)]
net: make use of internal/testenv package

Change-Id: I6644081df495cb92b3d208f867066f9acb08946f
Reviewed-on: https://go-review.googlesource.com/22074
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet: make newLocalPacketListener handle network argument correcly
Mikio Hara [Thu, 14 Apr 2016 00:57:43 +0000 (09:57 +0900)]
net: make newLocalPacketListener handle network argument correcly

Change-Id: I41691134770d01805c19c0f84f8828b00b85de0c
Reviewed-on: https://go-review.googlesource.com/22058
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/internal/obj: change Link.Flag_shared to bool
Matthew Dempsky [Thu, 14 Apr 2016 01:41:59 +0000 (18:41 -0700)]
cmd/internal/obj: change Link.Flag_shared to bool

Change-Id: I9bda2ce6f45fb8292503f86d8f9f161601f222b7
Reviewed-on: https://go-review.googlesource.com/22053
Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
9 years agocmd/compile/internal/gc: change flags to bool where possible
Matthew Dempsky [Thu, 14 Apr 2016 01:37:18 +0000 (18:37 -0700)]
cmd/compile/internal/gc: change flags to bool where possible

Some of the Debug[x] flags are actually boolean too, but not all, so
they need to be handled separately.

While here, change some obj.Flagstr and obj.Flagint64 calls to
directly use flag.StringVar and flag.Int64Var instead.

Change-Id: Iccedf6fed4328240ee2257f57fe6d66688f237c4
Reviewed-on: https://go-review.googlesource.com/22052
Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
9 years agocmd/compile: use correct export function (fix debugFormat)
Robert Griesemer [Wed, 13 Apr 2016 23:57:23 +0000 (16:57 -0700)]
cmd/compile: use correct export function (fix debugFormat)

Tested with debugFormat enabled and running
(export GO_GCFLAGS=-newexport; sh all.bash).

Change-Id: If7d43e1e594ea43c644232b89e670f7abb6b003e
Reviewed-on: https://go-review.googlesource.com/22033
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agofmt: clarify that for %g precision determines number of significant digits
Rob Pike [Wed, 13 Apr 2016 18:47:25 +0000 (11:47 -0700)]
fmt: clarify that for %g precision determines number of significant digits

Documentation change only.

Fixes #15178.

Change-Id: I3c7d80ce9e668ac7515f7ebb9da80f3bd8e534d6
Reviewed-on: https://go-review.googlesource.com/22006
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile: don't export unneeded OAS, OASWB nodes
Robert Griesemer [Wed, 13 Apr 2016 20:17:30 +0000 (13:17 -0700)]
cmd/compile: don't export unneeded OAS, OASWB nodes

Also:
- "rewrite" node Op in exporter for some nodes instead of importer
- more comments

Change-Id: I809e6754d14987b28f1da9379951ffa2e690c2a7
Reviewed-on: https://go-review.googlesource.com/22008
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agosyscall: fix epoll_event struct for ppc64le/ppc64
Lynn Boger [Wed, 6 Apr 2016 16:07:12 +0000 (11:07 -0500)]
syscall:  fix epoll_event struct for ppc64le/ppc64

The existing epoll_event structure used by many of
the epoll_* syscalls was defined incorrectly
for use with ppc64le & ppc64 in the syscall
directory.  This resulted in the caller getting
incorrect information on return from these
syscalls.  This caused failures in fsnotify as
well as builds with upstream Docker.  The
structure is defined correctly in gccgo.

This adds a pad field that is expected for
these syscalls on ppc64le, ppc64.
Fixes #15135

Change-Id: If7e8ea9eb1d1ca5182c8dc0f935b334127341ffd
Reviewed-on: https://go-review.googlesource.com/21582
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

9 years agocmd/compile, etc: use name for type pkgPath
David Crawshaw [Thu, 31 Mar 2016 14:02:10 +0000 (10:02 -0400)]
cmd/compile, etc: use name for type pkgPath

By replacing the *string used to represent pkgPath with a
reflect.name everywhere, the embedded *string for package paths
inside the reflect.name can be replaced by an offset, nameOff.
This reduces the number of pointers in the type information.

This also moves all reflect.name types into the same section, making
it possible to use nameOff more widely in later CLs.

No significant binary size change for normal binaries, but:

linux/amd64 PIE:
cmd/go: -440KB (3.7%)
jujud:  -2.6MB (3.2%)

For #6853.

Change-Id: I3890b132a784a1090b1b72b32febfe0bea77eaee
Reviewed-on: https://go-review.googlesource.com/21395
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: rename os1_darwin.go to os_darwin.go
Brad Fitzpatrick [Wed, 13 Apr 2016 18:33:42 +0000 (11:33 -0700)]
runtime: rename os1_darwin.go to os_darwin.go

Change-Id: If0e0bc5a85101db1e70faaab168fc2d12024eb93
Reviewed-on: https://go-review.googlesource.com/22005
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: merge the darwin os*.go files together
Brad Fitzpatrick [Wed, 13 Apr 2016 18:31:24 +0000 (11:31 -0700)]
runtime: merge the darwin os*.go files together

Merge them together into os1_darwin.go. A future CL will rename it.

Change-Id: Ia4380d3296ebd5ce210908ce3582ff184566f692
Reviewed-on: https://go-review.googlesource.com/22004
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/link: handle long symbol names
David Crawshaw [Wed, 6 Apr 2016 11:11:24 +0000 (07:11 -0400)]
cmd/link: handle long symbol names

Fixes #15104.

Change-Id: I9ddfbbf39ef0a873b703ee3e04fbb7d1192f5f39
Reviewed-on: https://go-review.googlesource.com/21581
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile: sort partitions by dom to speed up cse
Todd Neal [Wed, 13 Apr 2016 12:51:46 +0000 (08:51 -0400)]
cmd/compile: sort partitions by dom to speed up cse

We do two O(n) scans of all values in an eqclass when computing
substitutions for CSE.

In unfortunate cases, like those found in #15112, we can have a large
eqclass composed of values found in blocks none of whom dominate the
other.  This leads to O(n^2) behavior. The elements are removed one at a
time, with O(n) scans each time.

This CL removes the linear scan by sorting the eqclass so that dominant
values will be sorted first.  As long as we also ensure we don't disturb
the sort order, then we no longer need to scan for the maximally
dominant value.

For the code in issue #15112:

Before:
real    1m26.094s
user    1m30.776s
sys     0m1.125s

Aefter:
real    0m52.099s
user    0m56.829s
sys     0m1.092s

Updates #15112

Change-Id: Ic4f8680ed172e716232436d31963209c146ef850
Reviewed-on: https://go-review.googlesource.com/21981
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Todd Neal <todd@tneal.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime/internal/atomic: rename Storep1 to StorepNoWB
Austin Clements [Wed, 13 Apr 2016 15:22:42 +0000 (11:22 -0400)]
runtime/internal/atomic: rename Storep1 to StorepNoWB

Make it clear that the point of this function stores a pointer
*without* a write barrier.

sed -i -e 's/Storep1/StorepNoWB/' $(git grep -l Storep1)

Updates #15270.

Change-Id: Ifad7e17815e51a738070655fe3b178afdadaecf6
Reviewed-on: https://go-review.googlesource.com/21994
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
9 years agocmd/pprof/internal/profile: always subtract 1 from PCs
Dmitry Vyukov [Fri, 8 Apr 2016 15:56:36 +0000 (17:56 +0200)]
cmd/pprof/internal/profile: always subtract 1 from PCs

Go runtime never emits PCs that are not a return address
(except for cpu profiler).

Change-Id: I08d9dc5c7c71e23f34f2f0c16f8baeeb4f64fcd6
Reviewed-on: https://go-review.googlesource.com/21735
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile, go/importer: minor cleanups
Robert Griesemer [Wed, 13 Apr 2016 04:58:44 +0000 (21:58 -0700)]
cmd/compile, go/importer: minor cleanups

Change-Id: Ic7a1fb0dbbf108052c970a4a830269a5673df7df
Reviewed-on: https://go-review.googlesource.com/21963
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agocmd/compile: move more compiler specifics into compiler specific export section
Robert Griesemer [Wed, 13 Apr 2016 01:00:04 +0000 (18:00 -0700)]
cmd/compile: move more compiler specifics into compiler specific export section

Instead of indicating with each function signature if it has an inlineable
body, collect all functions in order and export function bodies with function
index in platform-specific section.

Moves this compiler specific information out of the platform-independent
export data section, and removes an int value for all functions w/o body.
Also simplifies the code a bit.

Change-Id: I8b2d7299dbe81f2706be49ecfb9d9f7da85fd854
Reviewed-on: https://go-review.googlesource.com/21939
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agoruntime/internal/atomic: remove write barrier from Storep1 on s390x
Austin Clements [Wed, 13 Apr 2016 15:13:39 +0000 (11:13 -0400)]
runtime/internal/atomic: remove write barrier from Storep1 on s390x

atomic.Storep1 is not supposed to invoke a write barrier (that's what
atomicstorep is for), but currently does on s390x. This causes a panic
in runtime.mapzero when it tries to use atomic.Storep1 to store what's
actually a scalar.

Fix this by eliminating the write barrier from atomic.Storep1 on
s390x. Also add some documentation to atomicstorep to explain the
difference between these.

Fixes #15270.

Change-Id: I291846732d82f090a218df3ef6351180aff54e81
Reviewed-on: https://go-review.googlesource.com/21993
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Munday <munday@ca.ibm.com>
9 years agocmd/compile: move spills to loop exits when easy.
David Chase [Mon, 21 Mar 2016 15:32:04 +0000 (11:32 -0400)]
cmd/compile: move spills to loop exits when easy.

For call-free inner loops.

Revised statistics:
  85 inner loop spills sunk
 341 inner loop spills remaining
1162 inner loop spills that were candidates for sinking
     ended up completely register allocated
 119 inner loop spills could have been sunk were used in
     "shuffling" at the bottom of the loop.
   1 inner loop spill not sunk because the register assigned
     changed between def and exit,

 Understanding how to make an inner loop definition not be
 a candidate for from-memory shuffling (to force the shuffle
 code to choose some other value) should pick up some of the
 119 other spills disqualified for this reason.

 Modified the stats printing based on feedback from Austin.

Change-Id: If3fb9b5d5a028f42ccc36c4e3d9e0da39db5ca60
Reviewed-on: https://go-review.googlesource.com/21037
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: improve memmove performance ppc64,ppc64le
Lynn Boger [Wed, 13 Apr 2016 13:58:10 +0000 (08:58 -0500)]
runtime: improve memmove performance ppc64,ppc64le

This change improves the performance of memmove
on ppc64 & ppc64le mainly for moves >=32 bytes.
In addition, the test to detect backward moves
 was enhanced to avoid backward moves if source
and dest were in different types of storage, since
backward moves might not always be efficient.

Fixes #14507

The following shows some of the improvements from the test
in the runtime package:

BenchmarkMemmove32                   4229.56      4717.13      1.12x
BenchmarkMemmove64                   6156.03      7810.42      1.27x
BenchmarkMemmove128                  7521.69      12468.54     1.66x
BenchmarkMemmove256                  6729.90      18260.33     2.71x
BenchmarkMemmove512                  8521.59      18033.81     2.12x
BenchmarkMemmove1024                 9760.92      25762.61     2.64x
BenchmarkMemmove2048                 10241.00     29584.94     2.89x
BenchmarkMemmove4096                 10399.37     31882.31     3.07x

BenchmarkMemmoveUnalignedDst16       1943.69      2258.33      1.16x
BenchmarkMemmoveUnalignedDst32       3885.08      3965.81      1.02x
BenchmarkMemmoveUnalignedDst64       5121.63      6965.54      1.36x
BenchmarkMemmoveUnalignedDst128      7212.34      11372.68     1.58x
BenchmarkMemmoveUnalignedDst256      6564.52      16913.59     2.58x
BenchmarkMemmoveUnalignedDst512      8364.35      17782.57     2.13x
BenchmarkMemmoveUnalignedDst1024     9539.87      24914.72     2.61x
BenchmarkMemmoveUnalignedDst2048     9199.23      21235.11     2.31x
BenchmarkMemmoveUnalignedDst4096     10077.39     25231.99     2.50x

BenchmarkMemmoveUnalignedSrc32       3249.83      3742.52      1.15x
BenchmarkMemmoveUnalignedSrc64       5562.35      6627.96      1.19x
BenchmarkMemmoveUnalignedSrc128      6023.98      10200.84     1.69x
BenchmarkMemmoveUnalignedSrc256      6921.83      15258.43     2.20x
BenchmarkMemmoveUnalignedSrc512      8593.13      16541.97     1.93x
BenchmarkMemmoveUnalignedSrc1024     9730.95      22927.84     2.36x
BenchmarkMemmoveUnalignedSrc2048     9793.28      21537.73     2.20x
BenchmarkMemmoveUnalignedSrc4096     10132.96     26295.06     2.60x

Change-Id: I73af59970d4c97c728deabb9708b31ec7e01bdf2
Reviewed-on: https://go-review.googlesource.com/21990
Reviewed-by: Bill O'Farrell <billotosyr@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/link: use a switch for name prefix switching
David Crawshaw [Wed, 13 Apr 2016 14:06:12 +0000 (10:06 -0400)]
cmd/link: use a switch for name prefix switching

Minor cleanup.

Change-Id: I7574f58a7e55c2bb798ebe9c7c98d36b8c258fb8
Reviewed-on: https://go-review.googlesource.com/21982
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile, etc: store method tables as offsets
David Crawshaw [Mon, 28 Mar 2016 14:32:27 +0000 (10:32 -0400)]
cmd/compile, etc: store method tables as offsets

This CL introduces the typeOff type and a lookup method of the same
name that can turn a typeOff offset into an *rtype.

In a typical Go binary (built with buildmode=exe, pie, c-archive, or
c-shared), there is one moduledata and all typeOff values are offsets
relative to firstmoduledata.types. This makes computing the pointer
cheap in typical programs.

With buildmode=shared (and one day, buildmode=plugin) there are
multiple modules whose relative offset is determined at runtime.
We identify a type in the general case by the pair of the original
*rtype that references it and its typeOff value. We determine
the module from the original pointer, and then use the typeOff from
there to compute the final *rtype.

To ensure there is only one *rtype representing each type, the
runtime initializes a typemap for each module, using any identical
type from an earlier module when resolving that offset. This means
that types computed from an offset match the type mapped by the
pointer dynamic relocations.

A series of followup CLs will replace other *rtype values with typeOff
(and name/*string with nameOff).

For types created at runtime by reflect, type offsets are treated as
global IDs and reference into a reflect offset map kept by the runtime.

darwin/amd64:
cmd/go:  -57KB (0.6%)
jujud:  -557KB (0.8%)

linux/amd64 PIE:
cmd/go: -361KB (3.0%)
jujud:  -3.5MB (4.2%)

For #6853.

Change-Id: Icf096fd884a0a0cb9f280f46f7a26c70a9006c96
Reviewed-on: https://go-review.googlesource.com/21285
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: use shared dom tree for cse, too
Alexandru Moșoi [Wed, 13 Apr 2016 08:58:38 +0000 (10:58 +0200)]
cmd/compile: use shared dom tree for cse, too

Missed this in the previous CL where the shared
dom tree was introduced.

Change-Id: If0bd85d4b4567d7e87814ed511603b1303ab3903
Reviewed-on: https://go-review.googlesource.com/21970
Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
9 years agocmd/link: rename Pcln to FuncInfo
Shahar Kohanim [Tue, 12 Apr 2016 20:18:47 +0000 (23:18 +0300)]
cmd/link: rename Pcln to FuncInfo

After non pcln fields were added to it in a previous commit.

Change-Id: Icf92c0774d157c61399a6fc2a3c4d2cd47a634d2
Reviewed-on: https://go-review.googlesource.com/21921
Run-TryBot: Shahar Kohanim <skohanim@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agocmd/compile: make enqueued map keys fail validation on forward types
Tal Shprecher [Mon, 11 Apr 2016 01:12:41 +0000 (18:12 -0700)]
cmd/compile: make enqueued map keys fail validation on forward types

Map keys are currently validated in multiple locations but share
a common validation routine. The problem is that early validations
should be lenient enough to allow for forward types while the final
validations should not. The final validations should fail on forward
types since they've already settled.

This change also separates the key type checking from the creation
of the map via typMap. Instead of the mapqueue being populated in
copytype() by checking the map line number, it's populated in the
same block that validates the key type. This isolates key validation
logic while type checking.

Fixes #14988

Change-Id: Ia47cf6213585d6c63b3a35249104c0439feae658
Reviewed-on: https://go-review.googlesource.com/21830
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: fix crash on bare package name in constant declarations
Matthew Dempsky [Wed, 13 Apr 2016 00:46:41 +0000 (17:46 -0700)]
cmd/compile: fix crash on bare package name in constant declarations

Fixes #11361.

Change-Id: I70b8808f97f0e07de680e7e6ede1322ea0fdbbc0
Reviewed-on: https://go-review.googlesource.com/21936
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: remove unnecessary assignments while type checking.
Tal Shprecher [Wed, 13 Apr 2016 05:29:34 +0000 (22:29 -0700)]
cmd/compile: remove unnecessary assignments while type checking.

Change-Id: Ica0ec84714d7f01d800d62fa10cdb08321d43cf3
Reviewed-on: https://go-review.googlesource.com/21967
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agoruntime: simplify setPanicOnFault slightly
Matthew Dempsky [Tue, 12 Apr 2016 22:51:24 +0000 (15:51 -0700)]
runtime: simplify setPanicOnFault slightly

No need to acquire the M just to change G's paniconfault flag, and the
original C implementation of SetPanicOnFault did not. The M
acquisition logic is an artifact of golang.org/cl/131010044, which was
started before golang.org/cl/123640043 (which introduced the current
"getg" function) was submitted.

Change-Id: I6d1939008660210be46904395cf5f5bbc2c8f754
Reviewed-on: https://go-review.googlesource.com/21935
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoall: use new io.SeekFoo constants instead of os.SEEK_FOO
Brad Fitzpatrick [Wed, 13 Apr 2016 04:35:37 +0000 (04:35 +0000)]
all: use new io.SeekFoo constants instead of os.SEEK_FOO

Automated change.

Fixes #15269

Change-Id: I8deb2ac0101d3f7c390467ceb0a1561b72edbb2f
Reviewed-on: https://go-review.googlesource.com/21962
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agostrings: improve explode and correct comment
Martin Möhrmann [Fri, 25 Mar 2016 23:04:48 +0000 (00:04 +0100)]
strings: improve explode and correct comment

Merges explodetests into splittests which already contain
some of the tests that cover explode.

Adds a test to cover the utf8.RuneError branch in explode.

name      old time/op  new time/op  delta
Split1-2  14.9ms ± 0%  14.2ms ± 0%  -4.06%  (p=0.000 n=47+49)

Change-Id: I00f796bd2edab70e926ea9e65439d820c6a28254
Reviewed-on: https://go-review.googlesource.com/21609
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agohtml/template: add examples of loading templates from files
Emmanuel Odeke [Tue, 12 Apr 2016 08:55:14 +0000 (01:55 -0700)]
html/template: add examples of loading templates from files

Adds examples showing loading templates from files and
executing them.

Shows examples:
- Using ParseGlob.
- Using ParseFiles.
- Using helper functions to share and use templates
in different contexts by adding them to an existing
bundle of templates.
- Using a group of driver templates with distinct sets
of helper templates.

Almost all of the code was directly copied from text/template.

Fixes #8500

Change-Id: Ic3d91d5232afc5a1cd2d8cd3d9a5f3b754c64225
Reviewed-on: https://go-review.googlesource.com/21854
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/compile: teach CSE that new objects are bespoke
Josh Bleecher Snyder [Wed, 13 Apr 2016 00:12:26 +0000 (17:12 -0700)]
cmd/compile: teach CSE that new objects are bespoke

runtime.newobject never returns the same thing twice,
so the resulting value will never be a common subexpression.

This helps when compiling large static data structures
that include pointers, such as maps and slices.
No clear performance impact on other code. (See below.)

For the code in issue #15112:

Before:
  real 1m14.238s
  user 1m18.985s
  sys 0m0.787s

After:
  real 0m47.172s
  user 0m52.248s
  sys 0m0.767s

For the code in issue #15235, size 10k:

Before:
  real 0m44.916s
  user 0m46.577s
  sys 0m0.304s

After:
  real 0m7.703s
  user 0m9.041s
  sys 0m0.316s

Still more work to be done, particularly for #15112.

Updates #15112
Updates #15235

name       old time/op      new time/op      delta
Template        330ms ±11%       333ms ±13%    ~           (p=0.749 n=20+19)
Unicode         148ms ± 6%       152ms ± 8%    ~           (p=0.072 n=18+20)
GoTypes         1.01s ± 7%       1.01s ± 3%    ~           (p=0.583 n=20+20)
Compiler        5.04s ± 2%       5.06s ± 2%    ~           (p=0.314 n=20+20)

name       old user-ns/op   new user-ns/op   delta
Template   444user-ms ±11%  445user-ms ±10%    ~           (p=0.738 n=20+20)
Unicode    215user-ms ± 5%  218user-ms ± 5%    ~           (p=0.239 n=18+18)
GoTypes    1.45user-s ± 3%  1.45user-s ± 4%    ~           (p=0.620 n=20+20)
Compiler   7.23user-s ± 2%  7.22user-s ± 2%    ~           (p=0.901 n=20+19)

name       old alloc/op     new alloc/op     delta
Template       55.0MB ± 0%      55.0MB ± 0%    ~           (p=0.547 n=20+20)
Unicode        37.6MB ± 0%      37.6MB ± 0%    ~           (p=0.301 n=20+20)
GoTypes         177MB ± 0%       177MB ± 0%    ~           (p=0.065 n=20+19)
Compiler        798MB ± 0%       797MB ± 0%  -0.05%        (p=0.000 n=19+20)

name       old allocs/op    new allocs/op    delta
Template         492k ± 0%        493k ± 0%  +0.03%        (p=0.030 n=20+20)
Unicode          377k ± 0%        377k ± 0%    ~           (p=0.423 n=20+19)
GoTypes         1.40M ± 0%       1.40M ± 0%    ~           (p=0.102 n=20+20)
Compiler        5.53M ± 0%       5.53M ± 0%    ~           (p=0.094 n=17+18)

name       old text-bytes   new text-bytes   delta
HelloSize        561k ± 0%        561k ± 0%    ~     (all samples are equal)
CmdGoSize       6.13M ± 0%       6.13M ± 0%    ~     (all samples are equal)

name       old data-bytes   new data-bytes   delta
HelloSize        128k ± 0%        128k ± 0%    ~     (all samples are equal)
CmdGoSize        306k ± 0%        306k ± 0%    ~     (all samples are equal)

name       old exe-bytes    new exe-bytes    delta
HelloSize        905k ± 0%        905k ± 0%    ~     (all samples are equal)
CmdGoSize       9.64M ± 0%       9.64M ± 0%    ~     (all samples are equal)

Change-Id: Id774e2901d7701a3ec7a1c1d1cf1d9327a4107fc
Reviewed-on: https://go-review.googlesource.com/21937
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Todd Neal <todd@tneal.org>
9 years agocmd/compile: fix arg to getcallerpc
Keith Randall [Tue, 12 Apr 2016 23:25:48 +0000 (16:25 -0700)]
cmd/compile: fix arg to getcallerpc

getcallerpc's arg needs to point to the first argument slot.
I believe this bug was introduced by Michel's itab changes
(specifically https://go-review.googlesource.com/c/20902).

Fixes #15145

Change-Id: Ifb2e17f3658e2136c7950dfc789b4d5706320683
Reviewed-on: https://go-review.googlesource.com/21931
Reviewed-by: Michel Lespinasse <walken@google.com>
9 years agocmd/link: move function only lsym fields to pcln struct
Shahar Kohanim [Mon, 11 Apr 2016 19:19:34 +0000 (22:19 +0300)]
cmd/link: move function only lsym fields to pcln struct

name       old secs    new secs    delta
LinkCmdGo   0.53 ± 9%   0.53 ±10%  -1.30%  (p=0.022 n=100+99)

name       old MaxRSS  new MaxRSS  delta
LinkCmdGo   151k ± 4%   142k ± 6%  -5.92%  (p=0.000 n=98+100)

Change-Id: Ic30e63a948f8e626b3396f458a0163f7234810c1
Reviewed-on: https://go-review.googlesource.com/21920
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet/http: add test that panic in a handler signals an error to the client
Brad Fitzpatrick [Tue, 12 Apr 2016 00:22:39 +0000 (00:22 +0000)]
net/http: add test that panic in a handler signals an error to the client

Change-Id: Iba40edc9ddad62534b06c5af20bbc3dd3dc14d0a
Reviewed-on: https://go-review.googlesource.com/21881
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoreflect: test that Call results are not addressable
Ian Lance Taylor [Tue, 12 Apr 2016 22:47:17 +0000 (15:47 -0700)]
reflect: test that Call results are not addressable

Gccgo was erroneously marking Call results as addressable, which led to
an obscure bug using text/template, as text/template calls CanAddr to
check whether to take the address of a value when looking up methods.
When a function returned a pointer, and CanAddr was true, the result was
a pointer to a pointer that had no methods.

Fixed in gccgo by https://golang.org/cl/21908.  Adding the test here so
that it doesn't regress.

Change-Id: I1d25b868e1b8e2348b21cbac6404a636376d1a4a
Reviewed-on: https://go-review.googlesource.com/21930
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoimage/color: optimize YCbCrToRGB
Martin Möhrmann [Tue, 12 Apr 2016 19:16:27 +0000 (21:16 +0200)]
image/color: optimize YCbCrToRGB

Use one comparison to detect underflow and overflow simultaneously.
Use a shift, bitwise complement and uint8 type conversion to handle
clamping to upper and lower bound without additional branching.

Overall the new code is faster for a mix of
common case, underflow and overflow.

name     old time/op  new time/op  delta
YCbCr-2  1.12ms ± 0%  0.64ms ± 0%  -43.01%  (p=0.000 n=48+47)

name              old time/op  new time/op  delta
YCbCrToRGB/0-2    5.52ns ± 0%  5.77ns ± 0%  +4.48%  (p=0.000 n=50+49)
YCbCrToRGB/128-2  6.05ns ± 0%  5.52ns ± 0%  -8.69%  (p=0.000 n=39+50)
YCbCrToRGB/255-2  5.80ns ± 0%  5.77ns ± 0%  -0.58%  (p=0.000 n=50+49)

Found in collaboration with Josh Bleecher Snyder and Ralph Corderoy.

Change-Id: Ic5020320f704966f545fdc1ae6bc24ddb5d3d09a
Reviewed-on: https://go-review.googlesource.com/21910
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: skip failing or flaky TestInterfaces on freebsd-arm
Brad Fitzpatrick [Tue, 12 Apr 2016 22:27:33 +0000 (22:27 +0000)]
net: skip failing or flaky TestInterfaces on freebsd-arm

Updates #15262

Change-Id: I3eb1f6f71d6285d039f11ba6a34b8a599a33bf49
Reviewed-on: https://go-review.googlesource.com/21909
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agonet: mirror Tom Sawyer in the net package for tests
Brad Fitzpatrick [Tue, 12 Apr 2016 21:51:19 +0000 (21:51 +0000)]
net: mirror Tom Sawyer in the net package for tests

Fixes the darwin/arm builder, which has a special test runner which
makes the assumption that tests never use testdata from another
package.

This looks large, but it's no more space in git.

Change-Id: I81921b516443d12d21b77617d323ddebedbe40f8
Reviewed-on: https://go-review.googlesource.com/21907
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/internal/obj: remove unused Pciter type
Matthew Dempsky [Tue, 12 Apr 2016 21:44:49 +0000 (14:44 -0700)]
cmd/internal/obj: remove unused Pciter type

Change-Id: Ie8323cfcd1193f390729d0d3dd67863aedf47d13
Reviewed-on: https://go-review.googlesource.com/21906
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: make two tests not parallel
Josh Bleecher Snyder [Tue, 12 Apr 2016 21:04:01 +0000 (14:04 -0700)]
net: make two tests not parallel

Running

stress -p 1 go test -short std

on a heavily loaded machine causes net timeouts
every 15 or 20 runs.
Making these tests not run in parallel helps.
With this change, I haven’t seen a single failure
in over 100 runs.

Fixes #14986

Change-Id: Ibaa14869ce8d95b00266aee94d62d195927ede68
Reviewed-on: https://go-review.googlesource.com/21905
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoall: standardize RFC mention format
Dan Peterson [Tue, 12 Apr 2016 19:58:56 +0000 (16:58 -0300)]
all: standardize RFC mention format

Standardize on space between "RFC" and number. Additionally change
the couple "a RFC" instances to "an RFC."

Fixes #15258

Change-Id: I2b17ecd06be07dfbb4207c690f52a59ea9b04808
Reviewed-on: https://go-review.googlesource.com/21902
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/internal/obj: split plist flushing from object writing
Shahar Kohanim [Mon, 11 Apr 2016 14:35:55 +0000 (17:35 +0300)]
cmd/internal/obj: split plist flushing from object writing

Only splits into separate files, no other changes.

Change-Id: Icc0da2c5f18e03e9ed7c0043bd7c790f741900f2
Reviewed-on: https://go-review.googlesource.com/21804
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/link, etc: store typelinks as offsets
David Crawshaw [Sun, 27 Mar 2016 14:21:48 +0000 (10:21 -0400)]
cmd/link, etc: store typelinks as offsets

This is the first in a series of CLs to replace the use of pointers
in binary read-only data with offsets.

In standard Go binaries these CLs have a small effect, shrinking
8-byte pointers to 4-bytes. In position-independent code, it also
saves the dynamic relocation for the pointer. This has a significant
effect on the binary size when building as PIE, c-archive, or
c-shared.

darwin/amd64:
cmd/go: -12KB (0.1%)
jujud:  -82KB (0.1%)

linux/amd64 PIE:
cmd/go:  -86KB (0.7%)
jujud:  -569KB (0.7%)

For #6853.

Change-Id: Iad5625bbeba58dabfd4d334dbee3fcbfe04b2dcf
Reviewed-on: https://go-review.googlesource.com/21284
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: skip TestDialCancel on linux-arm64-buildlet
Brad Fitzpatrick [Tue, 12 Apr 2016 19:48:01 +0000 (19:48 +0000)]
net: skip TestDialCancel on linux-arm64-buildlet

These builders (on Linaro) have a different network configuration
which is incompatible with this test. Or so it seems.

Updates #15191

Change-Id: Ibfeacddc98dac1da316e704b5c8491617a13e3bf
Reviewed-on: https://go-review.googlesource.com/21901
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agocmd/pprof: pass the event to pprof_toggle_asm for the weblist command
Alberto Donizetti [Sun, 10 Apr 2016 18:14:27 +0000 (20:14 +0200)]
cmd/pprof: pass the event to pprof_toggle_asm for the weblist command

Fixes #15225

Change-Id: I1f85590b2c3293463c6476beebcd3256adc1bf23
Reviewed-on: https://go-review.googlesource.com/21802
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocrypto/x509: remove broken link in ParsePKCS8PrivateKey documentation
Dan Peterson [Tue, 12 Apr 2016 16:12:54 +0000 (13:12 -0300)]
crypto/x509: remove broken link in ParsePKCS8PrivateKey documentation

Fixes #14776

Change-Id: I55423ac643f18542b9fd1386ed98dec47fb678aa
Reviewed-on: https://go-review.googlesource.com/21890
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd: replace x[i:][0] expressions with x[i]
Matthew Dempsky [Tue, 12 Apr 2016 19:16:20 +0000 (12:16 -0700)]
cmd: replace x[i:][0] expressions with x[i]

Passes toolstash -cmp.

Change-Id: Id504e71ed1f23900e24a9aed25143c94f4d7d50c
Reviewed-on: https://go-review.googlesource.com/21899
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: broken sendfile on SmartOS/Solaris
Shawn Walker-Salas [Fri, 8 Apr 2016 22:59:04 +0000 (15:59 -0700)]
net: broken sendfile on SmartOS/Solaris

In the event of a partial write on Solaris and some BSDs, the offset
pointer passed to sendfile() will be updated even though the function
returns -1 if errno is set to EAGAIN/EINTR.  In that case, calculate the
bytes written based on the difference between the updated offset and the
original offset.  If no bytes were written, and errno is set to
EAGAIN/EINTR, ignore the errno.

Fixes #13892

Change-Id: I6334b5ef2edcbebdaa7db36fa4f7785967313c2d
Reviewed-on: https://go-review.googlesource.com/21769
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: move compiler-specific flags into compiler-spec. export data section
Robert Griesemer [Tue, 12 Apr 2016 18:31:16 +0000 (11:31 -0700)]
cmd/compile: move compiler-specific flags into compiler-spec. export data section

Also: Adjust go/importer accordingly.

Change-Id: Ia6669563793e218946af45b9fba1cf986a21c031
Reviewed-on: https://go-review.googlesource.com/21896
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agocmd/objdump: skip TestDisasm* on s390x
Michael Munday [Tue, 12 Apr 2016 17:38:24 +0000 (13:38 -0400)]
cmd/objdump: skip TestDisasm* on s390x

The disassembler is not yet implemented on s390x.

Updates #15255.

Change-Id: Ibab319c8c087b1a619baa1529398305c1e721877
Reviewed-on: https://go-review.googlesource.com/21894
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/http: fix TestLinuxSendfile on s390x
Michael Munday [Tue, 12 Apr 2016 16:46:54 +0000 (12:46 -0400)]
net/http: fix TestLinuxSendfile on s390x

s390x doesn't have sendfile64 so apply the same fix as MIPS
(eebf7d27) and just use sendfile.

Change-Id: If8fe2e974ed44a9883282430157c3545d5bd04bd
Reviewed-on: https://go-review.googlesource.com/21892
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile/internal/gc: add s390x support
Michael Munday [Tue, 12 Apr 2016 16:26:17 +0000 (12:26 -0400)]
cmd/compile/internal/gc: add s390x support

Allows instructions with a From3 field to be used in regopt so
long as From3 represents a constant. This is needed because the
storage-to-storage instructions on s390x place the length of the
data into From3.

Change-Id: I12cd32d4f997baf2fe97937bb7d45bbf716dfcb5
Reviewed-on: https://go-review.googlesource.com/20875
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>

9 years agocmd/compile: temporarily disable inplace append special case
Josh Bleecher Snyder [Tue, 12 Apr 2016 16:22:26 +0000 (09:22 -0700)]
cmd/compile: temporarily disable inplace append special case

Fixes #15246
Re-opens #14969

Change-Id: Ic0b41c5aa42bbb229a0d62b7f3e5888c6b29293d
Reviewed-on: https://go-review.googlesource.com/21891
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agohash/crc32: invert build tags for go implementation
Michael Munday [Fri, 8 Apr 2016 17:30:41 +0000 (13:30 -0400)]
hash/crc32: invert build tags for go implementation

It seems cleaner and more consistent with other files to list the
architectures that have assembly implementations rather than to
list those that do not.

This means we don't have to add s390x and future platforms to this
list.

Change-Id: I2ad3f66b76eb1711333c910236ca7f5151b698e5
Reviewed-on: https://go-review.googlesource.com/21770
Reviewed-by: Bill O'Farrell <billotosyr@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime/cgo: add s390x support
Michael Munday [Fri, 18 Mar 2016 23:20:34 +0000 (19:20 -0400)]
runtime/cgo: add s390x support

Change-Id: I64ada9fe34c3cfc4bd514ec5d8c8f4d4c99074fb
Reviewed-on: https://go-review.googlesource.com/20950
Reviewed-by: Bill O'Farrell <billotosyr@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agodebug/gosym: accept PC quantum of 2 (for s390x)
Michael Munday [Mon, 21 Mar 2016 17:30:50 +0000 (13:30 -0400)]
debug/gosym: accept PC quantum of 2 (for s390x)

Needed for the header check to accept the header generated for
s390x as Go 1.2 style rather than Go 1.1 style.

Change-Id: I7b3713d4cc7514cfc58f947a45702348f6d7b824
Reviewed-on: https://go-review.googlesource.com/20966
Reviewed-by: Minux Ma <minux@golang.org>
9 years agoruntime/internal/atomic: add s390x atomic operations
Michael Munday [Fri, 18 Mar 2016 23:13:59 +0000 (19:13 -0400)]
runtime/internal/atomic: add s390x atomic operations

Load and store instructions are atomic on the s390x.

Change-Id: I0031ed2fba43f33863bca114d0fdec2e7d1ce807
Reviewed-on: https://go-review.googlesource.com/20938
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile/internal/gc: minor Cgen_checknil cleanup
Michael Munday [Tue, 12 Apr 2016 14:27:16 +0000 (10:27 -0400)]
cmd/compile/internal/gc: minor Cgen_checknil cleanup

Most architectures can only generate nil checks when the
the address to check is in a register. Currently only
amd64 and 386 can generate checks for addresses that
reside in memory. This is unlikely to change so the architecture
check has been inverted.

Change-Id: I73697488a183406c79a9039c62823712b510bb6a
Reviewed-on: https://go-review.googlesource.com/21861
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: add x.Uses==1 test to load combiners
Keith Randall [Tue, 12 Apr 2016 04:23:11 +0000 (21:23 -0700)]
cmd/compile: add x.Uses==1 test to load combiners

We need to make sure that when we combine loads, we only do
so if there are no other uses of the load.  We can't split
one load into two because that can then lead to inconsistent
loaded values in the presence of races.

Add some aggressive copy removal code so that phantom
"dead copy" uses of values are cleaned up promptly.  This lets
us use x.Uses==1 conditions reliably.

Change-Id: I9037311db85665f3868dbeb3adb3de5c20728b38
Reviewed-on: https://go-review.googlesource.com/21853
Reviewed-by: Todd Neal <todd@tneal.org>
9 years agocmd/pprof/internal: move to cmd/internal/pprof
Dmitry Vyukov [Tue, 12 Apr 2016 07:41:11 +0000 (09:41 +0200)]
cmd/pprof/internal: move to cmd/internal/pprof

Make internal pprof packages available to cmd/trace.
cmd/trace needs access to them to generate symbolized
svg profiles (create and serialize Profile struct).
And potentially generate svg programmatically instead
of invoking go tool pprof.

Change-Id: Iafd0c87ffdd4ddc081093be0b39761f19507907a
Reviewed-on: https://go-review.googlesource.com/21870
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: share dominator tree among many passes
Alexandru Moșoi [Mon, 11 Apr 2016 19:51:29 +0000 (21:51 +0200)]
cmd/compile: share dominator tree among many passes

These passes do not modify the dominator tree too much.

% benchstat old.txt new.txt
name       old time/op     new time/op     delta
Template       335ms ± 3%      325ms ± 8%    ~             (p=0.074 n=8+9)
GoTypes        1.05s ± 1%      1.05s ± 3%    ~            (p=0.095 n=9+10)
Compiler       5.37s ± 4%      5.29s ± 1%  -1.42%         (p=0.022 n=9+10)
MakeBash       34.9s ± 3%      34.4s ± 2%    ~            (p=0.095 n=9+10)

name       old alloc/op    new alloc/op    delta
Template      55.4MB ± 0%     54.9MB ± 0%  -0.81%        (p=0.000 n=10+10)
GoTypes        179MB ± 0%      178MB ± 0%  -0.89%        (p=0.000 n=10+10)
Compiler       807MB ± 0%      798MB ± 0%  -1.10%        (p=0.000 n=10+10)

name       old allocs/op   new allocs/op   delta
Template        498k ± 0%       496k ± 0%  -0.29%          (p=0.000 n=9+9)
GoTypes        1.42M ± 0%      1.41M ± 0%  -0.24%        (p=0.000 n=10+10)
Compiler       5.61M ± 0%      5.60M ± 0%  -0.12%        (p=0.000 n=10+10)

Change-Id: I4cd20cfba3f132ebf371e16046ab14d7e42799ec
Reviewed-on: https://go-review.googlesource.com/21806
Run-TryBot: Alexandru Moșoi <alexandru@mosoi.ro>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
9 years agocmd/compile/internal/s390x: add s390x support
Michael Munday [Tue, 12 Apr 2016 00:23:19 +0000 (20:23 -0400)]
cmd/compile/internal/s390x: add s390x support

s390x does not require duffzero/duffcopy since it has
storage-to-storage instructions that can copy/clear up to 256
bytes at a time.

peep contains several new passes to optimize instruction
sequences that match s390x instructions such as the
compare-and-branch and load/store multiple instructions.

copyprop and subprop have been extended to work with moves that
require sign/zero extension. This work could be ported to other
architectures that do not used sized math however it does add
complexity and will probably be rendered unnecessary by ssa in
the near future.

Change-Id: I1b64b281b452ed82a85655a0df69cb224d2a6941
Reviewed-on: https://go-review.googlesource.com/20873
Run-TryBot: Michael Munday <munday@ca.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bill O'Farrell <billotosyr@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agointernal/trace: fix int overflow in timestamps
Dmitry Vyukov [Fri, 8 Apr 2016 09:53:23 +0000 (11:53 +0200)]
internal/trace: fix int overflow in timestamps

Fixes #15102

Change-Id: I7fdb6464afd0b7af9b6652051416f0fddd34dc9a
Reviewed-on: https://go-review.googlesource.com/21730
Reviewed-by: Austin Clements <austin@google.com>
9 years agocmd/go: fix vcsFromDir returning bad root on Windows
Dmitri Shuralyov [Thu, 31 Mar 2016 09:01:48 +0000 (02:01 -0700)]
cmd/go: fix vcsFromDir returning bad root on Windows

Apply golang/tools@5804fef4c0556d3e5e7d2440740a3d7aced7d58b.

In the context of cmd/go build tool, import path is a '/'-separated path.
This can be inferred from `go help importpath` and `go help packages`.
vcsFromDir documentation says on return, root is the import path
corresponding to the root of the repository. On Windows and other
OSes where os.PathSeparator is not '/', that wasn't true since root
would contain characters other than '/', and therefore it wasn't a
valid import path corresponding to the root of the repository.
Fix that by using filepath.ToSlash.

Add test coverage for vcsFromDir, it was previously not tested.
It's taken from golang.org/x/tools/go/vcs tests, and modified to
improve style.

Additionally, remove an unneccessary statement from the documentation
"(thus root is a prefix of importPath)". There is no variable
importPath that is being referred to (it's possible p.ImportPath
was being referred to). Without it, the description of root value
matches the documentation of repoRoot.root struct field:

// root is the import path corresponding to the root of the
// repository
root string

Rename and change signature of vcsForDir(p *Package) to
vcsFromDir(dir, srcRoot string). This is more in sync with the x/tools
version. It's also simpler, since vcsFromDir only needs those two
values from Package, and nothing more. Change "for" to "from" in name
because it's more consistent and clear.

Update usage of vcsFromDir to match the new signature, and respect
that returned root is a '/'-separated path rather than a os.PathSeparator
separated path.

Fixes #15040.
Updates #7723.
Helps #11490.

Change-Id: Idf51b9239f57248739daaa200aa1c6e633cb5f7f
Reviewed-on: https://go-review.googlesource.com/21345
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: cleanup pragcgo
Martin Möhrmann [Thu, 7 Apr 2016 06:01:47 +0000 (08:01 +0200)]
cmd/compile: cleanup pragcgo

Removes dynimport, dynexport, dynlinker cases since they can not
be reached due to prefix check for "go:cgo_" in getlinepragma.

Replaces the if chains for verb distinction by a switch statement.
Replaces fmt.Sprintf by fmt.Sprintln for string concatenation.

Removes the more, getimpsym and getquoted functions by introducing a
pragmaFields function that partitions a pragma into its components.

Adds tests for cgo pragmas.

Change-Id: I43c7b9550feb3ddccaff7fb02198a3f994444123
Reviewed-on: https://go-review.googlesource.com/21607
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agosrc: split nacltest.bash into naclmake.bash and keep nacltest.bash
Brad Fitzpatrick [Tue, 12 Apr 2016 00:09:05 +0000 (00:09 +0000)]
src: split nacltest.bash into naclmake.bash and keep nacltest.bash

Needed by the build system to shard tests. nacl was the last unsharded
builder.

(I considered also adding a -make-only flag to nacltest.bash, but that
wouldn't fail fast when the file didn't exist.)

Updates #15242

Change-Id: I6afc1c1fe4268ab98c0724b5764c67d3784caebe
Reviewed-on: https://go-review.googlesource.com/21851
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocrypto/tls: Enforce that version and cipher match on resume.
David Benjamin [Mon, 15 Feb 2016 16:51:54 +0000 (11:51 -0500)]
crypto/tls: Enforce that version and cipher match on resume.

Per RFC 5246, 7.4.1.3:

   cipher_suite
      The single cipher suite selected by the server from the list in
      ClientHello.cipher_suites.  For resumed sessions, this field is
      the value from the state of the session being resumed.

The specifications are not very clearly written about resuming sessions
at the wrong version (i.e. is the TLS 1.0 notion of "session" the same
type as the TLS 1.1 notion of "session"?). But every other
implementation enforces this check and not doing so has some odd
semantics.

Change-Id: I6234708bd02b636c25139d83b0d35381167e5cad
Reviewed-on: https://go-review.googlesource.com/21153
Reviewed-by: Adam Langley <agl@golang.org>
9 years agoio: document WriteString calls Write exactly once
Dan Peterson [Tue, 12 Apr 2016 00:47:37 +0000 (21:47 -0300)]
io: document WriteString calls Write exactly once

Fixes #13849

Change-Id: Idd7f06b547a0179fe15571807a8c48b7c3b78d7c
Reviewed-on: https://go-review.googlesource.com/21852
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet: make IP.{String,MarshalText} return helpful information on address error
Mikio Hara [Thu, 7 Apr 2016 08:19:29 +0000 (17:19 +0900)]
net: make IP.{String,MarshalText} return helpful information on address error

This change makes String and MarshalText methods of IP return a
hexadecial form of IP with no punctuation as part of error
notification. It doesn't affect the existing behavior of ParseIP.

Also fixes bad shadowing in ipToSockaddr and makes use of reserved
IP address blocks for documnetation.

Fixes #15052.
Updates #15228.

Change-Id: I9e9ecce308952ed5683066c3d1bb6a7b36458c65
Reviewed-on: https://go-review.googlesource.com/21642
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agogo/importer: make For("gccgo", nil) not panic
Alan Donovan [Mon, 11 Apr 2016 22:39:46 +0000 (18:39 -0400)]
go/importer: make For("gccgo",  nil) not panic

Apparently we forgot to test this.

Fixes #15092

Change-Id: I33d4fef0f659dfbdfc1ebf8401e96610c8215592
Reviewed-on: https://go-review.googlesource.com/21860
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agocontext: attempt to deflake timing tests
Brad Fitzpatrick [Mon, 11 Apr 2016 18:16:55 +0000 (18:16 +0000)]
context: attempt to deflake timing tests

Passes on OpenBSD now when running it with -count=500.

Presumably this will also fix the same problems seen on FreeBSD and
Windows.

Fixes #15158

Change-Id: I86451c901613dfa5ecff0c2ecc516527a3c011b3
Reviewed-on: https://go-review.googlesource.com/21840
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/vet: improve documentation for flags, slightly
Rob Pike [Mon, 4 Apr 2016 20:22:34 +0000 (13:22 -0700)]
cmd/vet: improve documentation for flags, slightly

The way that -all works was unclear from the documentation and made
worse by recent changes to the flag package. Improve matters by making
the help message say "default true" for the tests that do default to true,
and tweak some of the wording.

Before:

Usage of vet:
vet [flags] directory...
vet [flags] files... # Must be a single package
For more information run
go doc cmd/vet

Flags:
  -all
     enable all non-experimental checks (default unset)
  -asmdecl
     check assembly against Go declarations (default unset)
...

After:

Usage of vet:
vet [flags] directory...
vet [flags] files... # Must be a single package
By default, -all is set and all non-experimental checks are run.
For more information run
go doc cmd/vet

Flags:
  -all
     enable all non-experimental checks (default true)
  -asmdecl
     check assembly against Go declarations (default true)
...

Change-Id: Ie94b27381a9ad2382a10a7542a93bce1d59fa8f5
Reviewed-on: https://go-review.googlesource.com/21495
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/compile: avoid write barrier in append fast path
Josh Bleecher Snyder [Sun, 10 Apr 2016 16:44:17 +0000 (09:44 -0700)]
cmd/compile: avoid write barrier in append fast path

When we are writing the result of an append back
to the same slice, we don’t need a write barrier
on the fast path.

This re-implements an optimization that was present
in the old backend.

Updates #14921
Fixes #14969

Sample code:

var x []byte

func p() {
x = append(x, 1, 2, 3)
}

Before:

"".p t=1 size=224 args=0x0 locals=0x48
0x0000 00000 (append.go:21) TEXT "".p(SB), $72-0
0x0000 00000 (append.go:21) MOVQ (TLS), CX
0x0009 00009 (append.go:21) CMPQ SP, 16(CX)
0x000d 00013 (append.go:21) JLS 199
0x0013 00019 (append.go:21) SUBQ $72, SP
0x0017 00023 (append.go:21) FUNCDATA $0, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
0x0017 00023 (append.go:21) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
0x0017 00023 (append.go:19) MOVQ "".x+16(SB), CX
0x001e 00030 (append.go:19) MOVQ "".x(SB), DX
0x0025 00037 (append.go:19) MOVQ "".x+8(SB), BX
0x002c 00044 (append.go:19) MOVQ BX, "".autotmp_0+64(SP)
0x0031 00049 (append.go:22) LEAQ 3(BX), BP
0x0035 00053 (append.go:22) CMPQ BP, CX
0x0038 00056 (append.go:22) JGT $0, 131
0x003a 00058 (append.go:22) MOVB $1, (DX)(BX*1)
0x003e 00062 (append.go:22) MOVB $2, 1(DX)(BX*1)
0x0043 00067 (append.go:22) MOVB $3, 2(DX)(BX*1)
0x0048 00072 (append.go:22) MOVQ BP, "".x+8(SB)
0x004f 00079 (append.go:22) MOVQ CX, "".x+16(SB)
0x0056 00086 (append.go:22) MOVL runtime.writeBarrier(SB), AX
0x005c 00092 (append.go:22) TESTB AL, AL
0x005e 00094 (append.go:22) JNE $0, 108
0x0060 00096 (append.go:22) MOVQ DX, "".x(SB)
0x0067 00103 (append.go:23) ADDQ $72, SP
0x006b 00107 (append.go:23) RET
0x006c 00108 (append.go:22) LEAQ "".x(SB), CX
0x0073 00115 (append.go:22) MOVQ CX, (SP)
0x0077 00119 (append.go:22) MOVQ DX, 8(SP)
0x007c 00124 (append.go:22) PCDATA $0, $0
0x007c 00124 (append.go:22) CALL runtime.writebarrierptr(SB)
0x0081 00129 (append.go:23) JMP 103
0x0083 00131 (append.go:22) LEAQ type.[]uint8(SB), AX
0x008a 00138 (append.go:22) MOVQ AX, (SP)
0x008e 00142 (append.go:22) MOVQ DX, 8(SP)
0x0093 00147 (append.go:22) MOVQ BX, 16(SP)
0x0098 00152 (append.go:22) MOVQ CX, 24(SP)
0x009d 00157 (append.go:22) MOVQ BP, 32(SP)
0x00a2 00162 (append.go:22) PCDATA $0, $0
0x00a2 00162 (append.go:22) CALL runtime.growslice(SB)
0x00a7 00167 (append.go:22) MOVQ 40(SP), DX
0x00ac 00172 (append.go:22) MOVQ 48(SP), AX
0x00b1 00177 (append.go:22) MOVQ 56(SP), CX
0x00b6 00182 (append.go:22) ADDQ $3, AX
0x00ba 00186 (append.go:19) MOVQ "".autotmp_0+64(SP), BX
0x00bf 00191 (append.go:22) MOVQ AX, BP
0x00c2 00194 (append.go:22) JMP 58
0x00c7 00199 (append.go:22) NOP
0x00c7 00199 (append.go:21) CALL runtime.morestack_noctxt(SB)
0x00cc 00204 (append.go:21) JMP 0

After:

"".p t=1 size=208 args=0x0 locals=0x48
0x0000 00000 (append.go:21) TEXT "".p(SB), $72-0
0x0000 00000 (append.go:21) MOVQ (TLS), CX
0x0009 00009 (append.go:21) CMPQ SP, 16(CX)
0x000d 00013 (append.go:21) JLS 191
0x0013 00019 (append.go:21) SUBQ $72, SP
0x0017 00023 (append.go:21) FUNCDATA $0, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
0x0017 00023 (append.go:21) FUNCDATA $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
0x0017 00023 (append.go:19) MOVQ "".x+16(SB), CX
0x001e 00030 (append.go:19) MOVQ "".x+8(SB), DX
0x0025 00037 (append.go:19) MOVQ DX, "".autotmp_0+64(SP)
0x002a 00042 (append.go:19) MOVQ "".x(SB), BX
0x0031 00049 (append.go:22) LEAQ 3(DX), BP
0x0035 00053 (append.go:22) MOVQ BP, "".x+8(SB)
0x003c 00060 (append.go:22) CMPQ BP, CX
0x003f 00063 (append.go:22) JGT $0, 84
0x0041 00065 (append.go:22) MOVB $1, (BX)(DX*1)
0x0045 00069 (append.go:22) MOVB $2, 1(BX)(DX*1)
0x004a 00074 (append.go:22) MOVB $3, 2(BX)(DX*1)
0x004f 00079 (append.go:23) ADDQ $72, SP
0x0053 00083 (append.go:23) RET
0x0054 00084 (append.go:22) LEAQ type.[]uint8(SB), AX
0x005b 00091 (append.go:22) MOVQ AX, (SP)
0x005f 00095 (append.go:22) MOVQ BX, 8(SP)
0x0064 00100 (append.go:22) MOVQ DX, 16(SP)
0x0069 00105 (append.go:22) MOVQ CX, 24(SP)
0x006e 00110 (append.go:22) MOVQ BP, 32(SP)
0x0073 00115 (append.go:22) PCDATA $0, $0
0x0073 00115 (append.go:22) CALL runtime.growslice(SB)
0x0078 00120 (append.go:22) MOVQ 40(SP), CX
0x007d 00125 (append.go:22) MOVQ 56(SP), AX
0x0082 00130 (append.go:22) MOVQ AX, "".x+16(SB)
0x0089 00137 (append.go:22) MOVL runtime.writeBarrier(SB), AX
0x008f 00143 (append.go:22) TESTB AL, AL
0x0091 00145 (append.go:22) JNE $0, 168
0x0093 00147 (append.go:22) MOVQ CX, "".x(SB)
0x009a 00154 (append.go:22) MOVQ "".x(SB), BX
0x00a1 00161 (append.go:19) MOVQ "".autotmp_0+64(SP), DX
0x00a6 00166 (append.go:22) JMP 65
0x00a8 00168 (append.go:22) LEAQ "".x(SB), DX
0x00af 00175 (append.go:22) MOVQ DX, (SP)
0x00b3 00179 (append.go:22) MOVQ CX, 8(SP)
0x00b8 00184 (append.go:22) PCDATA $0, $0
0x00b8 00184 (append.go:22) CALL runtime.writebarrierptr(SB)
0x00bd 00189 (append.go:22) JMP 154
0x00bf 00191 (append.go:22) NOP
0x00bf 00191 (append.go:21) CALL runtime.morestack_noctxt(SB)
0x00c4 00196 (append.go:21) JMP 0

Change-Id: I77a41ad3a22557a4bb4654de7d6d24a029efe34a
Reviewed-on: https://go-review.googlesource.com/21813
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agocmd/compile: added stats printing to stackalloc
David Chase [Fri, 1 Apr 2016 18:51:29 +0000 (14:51 -0400)]
cmd/compile: added stats printing to stackalloc

This is controlled by the "regalloc" stats flag, since regalloc
calls stackalloc.  The plan is for this to allow comparison
of cheaper stack allocation algorithms with what we have now.

Change-Id: Ibf64a780344c69babfcbb328fd6d053ea2e02cfc
Reviewed-on: https://go-review.googlesource.com/21393
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agocmd/compile: zero all three argstorage slots
Keith Randall [Mon, 11 Apr 2016 20:17:52 +0000 (13:17 -0700)]
cmd/compile: zero all three argstorage slots

These changes were missed when going from 2 to 3 argstorage slots.
https://go-review.googlesource.com/20296/

Change-Id: I930a307bb0b695bf1ae088030c9bbb6d14ca31d2
Reviewed-on: https://go-review.googlesource.com/21841
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
9 years agocmd/compile: fix -N build
Keith Randall [Mon, 11 Apr 2016 19:22:26 +0000 (12:22 -0700)]
cmd/compile: fix -N build

The decomposer of builtin types is confused by having structs
still around from the user-type decomposer.  They're all dead though,
so just enabling a deadcode pass fixes things.

Change-Id: I2df6bc7e829be03eabfd24c8dda1bff96f3d7091
Reviewed-on: https://go-review.googlesource.com/21839
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agoimage/color: add YCbCrToRGB benchmark
Josh Bleecher Snyder [Mon, 11 Apr 2016 18:54:07 +0000 (11:54 -0700)]
image/color: add YCbCrToRGB benchmark

Change-Id: I9ba76d5b0861a901415fdceccaf2f5caa2facb7f
Reviewed-on: https://go-review.googlesource.com/21837
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/internal/obj/s390x: add MULHD instruction
Michael Munday [Mon, 11 Apr 2016 00:01:49 +0000 (20:01 -0400)]
cmd/internal/obj/s390x: add MULHD instruction

Emulate 64-bit signed high multiplication ((a*b)>>64). To do this
we use the 64-bit unsigned high multiplication method and then
fix the result as shown in Hacker's Delight 2nd ed., chapter 8-3.

Required to enable some division optimizations.

Change-Id: I9194f428e09d3d029cb1afb4715cd5424b5d922e
Reviewed-on: https://go-review.googlesource.com/21774
Reviewed-by: Bill O'Farrell <billotosyr@gmail.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/http: add ServerContextKey to let a handler access its Server
Brad Fitzpatrick [Mon, 11 Apr 2016 05:12:43 +0000 (05:12 +0000)]
net/http: add ServerContextKey to let a handler access its Server

Fixes #12438
Updates #15229 (to decide context key variable naming convention)

Change-Id: I3ba423e91b689e232143247d044495a12c97a7d2
Reviewed-on: https://go-review.googlesource.com/21829
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/link: external linking can fail on Solaris 11.2+
Shawn Walker-Salas [Thu, 7 Apr 2016 22:26:57 +0000 (15:26 -0700)]
cmd/link: external linking can fail on Solaris 11.2+

Workaround external linking issues encountered on Solaris 11.2+ due to
the go.o object file being created with a NULL STT_FILE symtab entry by
using a placeholder name.

Fixes #14957

Change-Id: I89c501b4c548469f3c878151947d35588057982b
Reviewed-on: https://go-review.googlesource.com/21636
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agointernal/trace: support parsing of 1.5 traces
Dmitry Vyukov [Mon, 11 Apr 2016 06:57:52 +0000 (08:57 +0200)]
internal/trace: support parsing of 1.5 traces

1. Parse out version from trace header.
2. Restore handling of 1.5 traces.
3. Restore optional symbolization of traces.
4. Add some canned 1.5 traces for regression testing
   (http benchmark trace, runtime/trace stress traces,
    plus one with broken timestamps).

Change-Id: Idb18a001d03ded8e13c2730eeeb37c5836e31256
Reviewed-on: https://go-review.googlesource.com/21803
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>