]> Cypherpunks repositories - gostls13.git/log
gostls13.git
7 years agocmd/go: add source file contents to plugin hash
David Crawshaw [Sat, 2 Sep 2017 14:31:52 +0000 (10:31 -0400)]
cmd/go: add source file contents to plugin hash

It is common to have multiple plugins built from ephemeral
source files all with the same name:

# generate main.go
go build -buildmode=plugin -o=p1.so main.go
# rm main.go, generate new main.go
go build -buildmode=plugin -o=p2.so main.go
...

These different plugins currently have the same build ID,
and hence the same package path. This means only one can be
loaded.

To remove this restriction, this commit adds the contents of the
main package source files to the plugin hash.

Fixes #19358

Change-Id: Icd42024b085feb29c09c2771aaecb85f8b528dd3
Reviewed-on: https://go-review.googlesource.com/61170
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/pack: fix export data truncation bug
Matthew Dempsky [Thu, 31 Aug 2017 18:26:28 +0000 (11:26 -0700)]
cmd/pack: fix export data truncation bug

The binary export data format includes escaping to prevent "\n$$" from
appearing internally, but not "\n!\n". This could result in a false
positive when cmd/pack searched for "\n!\n" as the delimiter between
package definition and linker object.

To address this, this CL changes cmd/pack to also be aware of the
"\n$$" markers, and to ignore "\n!\n" within the export data.

Fixes #21703.

Change-Id: I71ea8ba49dbd066c7afb7717ddc0190e38fe5649
Reviewed-on: https://go-review.googlesource.com/60773
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agotest: add test that caused gccgo incorrect compilation
Than McIntosh [Fri, 8 Sep 2017 15:11:19 +0000 (11:11 -0400)]
test: add test that caused gccgo incorrect compilation

Updates #21770

Change-Id: Ic31c3bdae30797f406f25c737b83bbe2de1ed1da
Reviewed-on: https://go-review.googlesource.com/62331
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agotime: change wording in duration hours example
Juan Carlos [Fri, 21 Jul 2017 22:55:57 +0000 (17:55 -0500)]
time: change wording in duration hours example

Change-Id: I86728a1c6c20471beaa3546ca7a43a8edeb9f0b7
Reviewed-on: https://go-review.googlesource.com/50691
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adrian Hesketh <adrianhesketh@hushmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/dist: move functions for the better
Marvin Stenger [Thu, 31 Aug 2017 11:00:19 +0000 (13:00 +0200)]
cmd/dist: move functions for the better

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (6).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I1c49e3427079194210a6416057100a7e94a37619
Reviewed-on: https://go-review.googlesource.com/61012
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/dist: remove trivial variables + functions
Marvin Stenger [Thu, 31 Aug 2017 10:44:28 +0000 (12:44 +0200)]
cmd/dist: remove trivial variables + functions

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (5).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I0efd1271b6a70bb9248d82f8a4d869556f4a557e
Reviewed-on: https://go-review.googlesource.com/61011
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/dist: use standard generated code header
Marvin Stenger [Thu, 31 Aug 2017 10:27:53 +0000 (12:27 +0200)]
cmd/dist: use standard generated code header

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (4).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I5e163f89a518f074e58bf2d44597e553c918d7e6
Reviewed-on: https://go-review.googlesource.com/61010
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/dist: unleash bootstrap optimization for windows
Marvin Stenger [Sun, 3 Sep 2017 18:01:00 +0000 (20:01 +0200)]
cmd/dist: unleash bootstrap optimization for windows

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (3).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I7612bbc3939e6fca3bee4b8e92c528178dd46cd7
Reviewed-on: https://go-review.googlesource.com/61023
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/dist: restore behavior of branchtag
Marvin Stenger [Sun, 3 Sep 2017 17:58:01 +0000 (19:58 +0200)]
cmd/dist: restore behavior of branchtag

The behavior was changed unintentionally during the conversion from C to Go.
Reviewed-on: https://go-review.googlesource.com/2470

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (2).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I53373c7211b35ed68da485c55e510871bfb81267
Reviewed-on: https://go-review.googlesource.com/61022
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/dist: apply minor fixes
Marvin Stenger [Thu, 31 Aug 2017 10:21:51 +0000 (12:21 +0200)]
cmd/dist: apply minor fixes

This belongs to a series of clean-up changes (see below) for cmd/dist.
This is change (1).

These changes include:
(1)  apply minor fixes
(2)  restore behavior of branchtag
(3)  unleash bootstrap optimization for windows
(4)  use standard generated code header
(5)  remove trivial variables + functions
(6)  move functions for the better
(7)  simplify code segments
(8)  use bytes.Buffer for code generation
(9)  rename variables + functions
(10) remove doc.go

Change-Id: I49e5f2a9b6146e2b60a067da5bac31434ffc9aaf
Reviewed-on: https://go-review.googlesource.com/60650
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/link: extract windows-specific code from dynrelocsym
Hiroshi Ioka [Sat, 9 Sep 2017 02:23:29 +0000 (11:23 +0900)]
cmd/link: extract windows-specific code from dynrelocsym

No functional changes.

Change-Id: Ib31bb3f01b515aac6428ec61e0ef02b269623890
Reviewed-on: https://go-review.googlesource.com/62470
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agotime: add an example to the NewTicker function
Elbert Fliek [Tue, 5 Sep 2017 18:21:41 +0000 (20:21 +0200)]
time: add an example to the NewTicker function

Change-Id: Idad9cdee36679373ee223ff2bd4c021ea0afcce1
Reviewed-on: https://go-review.googlesource.com/61710
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: slightly more idiomatic println code
Matthew Dempsky [Fri, 8 Sep 2017 23:25:25 +0000 (16:25 -0700)]
cmd/compile: slightly more idiomatic println code

Updates #21808.

Change-Id: I0314426afcfeed17b1111040110d7f2b0e209526
Reviewed-on: https://go-review.googlesource.com/62430
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoregexp: Remove duplicated function wordRune()
Sylvain Zimmer [Sat, 2 Sep 2017 15:38:45 +0000 (17:38 +0200)]
regexp: Remove duplicated function wordRune()

Fixes #21742

Change-Id: Ib56b092c490c27a4ba7ebdb6391f1511794710b8
Reviewed-on: https://go-review.googlesource.com/61034
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
7 years agomath/rand: change http to https in comment
Ian Lance Taylor [Fri, 8 Sep 2017 13:03:46 +0000 (06:03 -0700)]
math/rand: change http to https in comment

Change-Id: I19c1b0e1b238dda82e69bd47459528ed06b55840
Reviewed-on: https://go-review.googlesource.com/62310
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years agodoc: add a diagnostics guide
Jaana Burcu Dogan [Tue, 5 Sep 2017 20:15:56 +0000 (13:15 -0700)]
doc: add a diagnostics guide

Diagnostics guide lists various dimensions of
diagnostics tools and libraries available in Go.
As a follow-up, I will add an entry section where
we navigate user to the right tool depending on
the type of problem they are willing to improve
or understand better.

Change-Id: I4e94b4b834014f51c988103457da84200c7827d9
Reviewed-on: https://go-review.googlesource.com/61693
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
7 years agocmd/compile: fix println()
Keith Randall [Fri, 8 Sep 2017 18:47:16 +0000 (11:47 -0700)]
cmd/compile: fix println()

println with no arguments accidentally doesn't print a newline.

Introduced at CL 55097

Fixes #21808

Change-Id: I9fc7b4271b9b31e4c9b6078f055195dc3907b62c
Reviewed-on: https://go-review.googlesource.com/62390
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/internal/goobj: add tests
Hiroshi Ioka [Wed, 30 Aug 2017 02:14:40 +0000 (11:14 +0900)]
cmd/internal/goobj: add tests

Change-Id: I4a0fe1c8625e7e9adfd84ac6910da83d0268f928
Reviewed-on: https://go-review.googlesource.com/60130
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: propagate constants through math.Float{32,64}{,from}bits
Michael Munday [Fri, 8 Sep 2017 00:31:13 +0000 (01:31 +0100)]
cmd/compile: propagate constants through math.Float{32,64}{,from}bits

This CL adds generic SSA rules to propagate constants through raw bits
conversions between floats and integers. This allows constants to
propagate through some math functions. For example, math.Copysign(0, -1)
is now constant folded to a load of -0.0.

Requires a fix to the ARM assembler which loaded -0.0 as +0.0.

Change-Id: I52649a4691077c7414f19d17bb599a6743c23ac2
Reviewed-on: https://go-review.googlesource.com/62250
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agocmd/go: don't write file if -n option is given
Hiroshi Ioka [Sat, 2 Sep 2017 01:36:01 +0000 (10:36 +0900)]
cmd/go: don't write file if -n option is given

Change-Id: I01f5d3b4748d0ead8642ff3e53d1ae9c4378bcbc
Reviewed-on: https://go-review.googlesource.com/61111
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agodatabase/sql: fix Conn's doc typo
Davor Kapsa [Fri, 8 Sep 2017 10:59:59 +0000 (12:59 +0200)]
database/sql: fix Conn's doc typo

Fixes #21798

Change-Id: Ided31a8b22c220acdeb3938cac41ce8db9a110c3
Reviewed-on: https://go-review.googlesource.com/62290
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
7 years agomath/rand: make Perm match Shuffle
Josh Bleecher Snyder [Tue, 15 Aug 2017 23:01:31 +0000 (16:01 -0700)]
math/rand: make Perm match Shuffle

Perm and Shuffle are fundamentally doing the same work.
This change makes Perm's algorithm match Shuffle's.
In addition to allowing developers to switch more
easily between the two methods, it affords a nice speed-up:

name      old time/op  new time/op  delta
Perm3-8   75.7ns ± 1%  51.8ns ± 1%  -31.59%  (p=0.000 n=9+8)
Perm30-8   610ns ± 1%   405ns ± 1%  -33.67%  (p=0.000 n=9+9)

This change alters the output from Perm,
given the same Source and seed.
This is a change from Go 1.0 behavior.
This necessitates updating the regression test.

This also changes the number of calls made to the Source
during Perm, which changes the output of the math/rand examples.

This also slightly perturbs the output of Perm,
nudging it out of the range currently accepted by TestUniformFactorial.
However, it is complete unclear that the helpers relied on
by TestUniformFactorial are correct. That is #21211.
This change updates checkSimilarDistribution to respect
closeEnough for standard deviations, which makes the test pass.
The whole situation is muddy; see #21211 for details.

There is an alternative implementation of Perm
that avoids initializing m, which is more similar
to the existing implementation, plus some optimizations:

func (r *Rand) Perm(n int) []int {
m := make([]int, n)
max31 := n
if n > 1<<31-1-1 {
max31 = 1<<31 - 1 - 1
}
i := 1
for ; i < max31; i++ {
j := r.int31n(int32(i + 1))
m[i] = m[j]
m[j] = i
}
for ; i < n; i++ {
j := r.Int63n(int64(i + 1))
m[i] = m[j]
m[j] = i
}
return m
}

This is a tiny bit faster than the implementation
actually used in this change:

name      old time/op  new time/op  delta
Perm3-8   51.8ns ± 1%  50.3ns ± 1%  -2.83%  (p=0.000 n=8+9)
Perm30-8   405ns ± 1%   394ns ± 1%  -2.66%  (p=0.000 n=9+8)

However, 3% in performance doesn't seem worth
having the two algorithms diverge,
nor the reduced readability of this alternative.

Updates #16213.

Change-Id: I11a7441ff8837ee9c241b4c88f7aa905348be781
Reviewed-on: https://go-review.googlesource.com/55972
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
7 years agomath/rand: add Shuffle
Josh Bleecher Snyder [Wed, 26 Jul 2017 00:53:30 +0000 (17:53 -0700)]
math/rand: add Shuffle

Shuffle uses the Fisher-Yates algorithm.

Since this is new API, it affords us the opportunity
to use a much faster Int31n implementation that mostly avoids division.
As a result, BenchmarkPerm30ViaShuffle is
about 30% faster than BenchmarkPerm30,
despite requiring a separate initialization loop
and using function calls to swap elements.

Fixes #20480
Updates #16213
Updates #21211

Change-Id: Ib8956c4bebed9d84f193eb98282ec16ee7c2b2d5
Reviewed-on: https://go-review.googlesource.com/51891
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agotime: don't match '---' month in time.Parse
Cholerae Hu [Tue, 25 Jul 2017 06:04:50 +0000 (14:04 +0800)]
time: don't match '---' month in time.Parse

The existing implementation will panic when month in date string is '---'.

Fixed #21113

Change-Id: I8058ae7a4102e882f8b7e9c65d80936b563265e4
Reviewed-on: https://go-review.googlesource.com/51010
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agopath: add path.Dir example with trailing slash
Mayank Kumar [Fri, 8 Sep 2017 08:04:28 +0000 (01:04 -0700)]
path: add path.Dir example with trailing slash

Change-Id: I143203a9dcf9a4da0e53a3aab6e370244b849296
Reviewed-on: https://go-review.googlesource.com/62270
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agonet/http: make ServeMux preserve query string during redirects
Kunpei Sakai [Sat, 2 Sep 2017 17:08:02 +0000 (02:08 +0900)]
net/http: make ServeMux preserve query string during redirects

Ensure that the implicitly created redirect
for
  "/route"
after
  "/route/"
has been registered doesn't lose the query string information.
A previous attempt (https://golang.org/cl/43779) changed http.Redirect, however, that change broke direct calls to http.Redirect.
To avoid that problem, this change touches ServeMux.Handler only.

Fixes #17841

Change-Id: I303c1b1824615304ae68147e254bb41b0ea339be
Reviewed-on: https://go-review.googlesource.com/61210
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tom Bergan <tombergan@google.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
7 years agocmd/asm: restrict x86 shift ops to 8bit args
isharipo [Thu, 7 Sep 2017 10:30:43 +0000 (13:30 +0300)]
cmd/asm: restrict x86 shift ops to 8bit args

Change "yshl" and "yshb" immediate oclass from Yi32 to Yu8.
This forbids:
- negative shift counts
- shift counts that not fit into 8bit

Affects:
  RCL{B,L,Q,W}
  RCR{B,L,Q,W}
  ROL{B,L,Q,W}
  ROR{B,L,Q,W}
  SAL{B,L,Q,W}
  SAR{B,L,Q,W}
  SHL{B,L,Q,W}
  SHR{B,L,Q,W}

Issue #21528 has some additional context about this change.

Change-Id: I60884cb2b41a860820889fcd878ca6f564006b4a
Reviewed-on: https://go-review.googlesource.com/62190
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agotext/template: fix pos info when trimming newlines
Daniel Martí [Wed, 6 Sep 2017 09:57:16 +0000 (11:57 +0200)]
text/template: fix pos info when trimming newlines

The lexer keeps the byte offset and the line for the rune it's currently
on. This was simple enough up until whitespace trimming was introduced.

With whitespace trimming, we might skip over newlines. In that case, the
lexer wasn't properly updating the line counter. Fix it.

Also, TestPos now checks that the line is correct too, which it was
ignoring before. This was necessary to test this scenario in the lexer.

Fixes #21778.

Change-Id: I3880f3adf02662eac8f818d5caa6935cca9cb33b
Reviewed-on: https://go-review.googlesource.com/61870
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Rob Pike <r@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/vet: remove two unused parameters and simplify
Daniel Martí [Wed, 6 Sep 2017 19:59:00 +0000 (21:59 +0200)]
cmd/vet: remove two unused parameters and simplify

The isStar and directory function parameters have been unused ever since
they were introduced. Remove them.

While at it, apply some other minor simplifications, such as simplifying
a HasPrefix if and using an early continue to unindent many lines of
code.

Change-Id: I8d57353e9ec10cdb59c5388cf6152ce0ec17bdba
Reviewed-on: https://go-review.googlesource.com/62030
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Marvin Stenger <marvin.stenger94@gmail.com>
7 years agoruntime: avoid extra tophash check in mapassign when key comparison is cheap
Martin Möhrmann [Wed, 6 Sep 2017 06:36:10 +0000 (08:36 +0200)]
runtime: avoid extra tophash check in mapassign when key comparison is cheap

mapaccess and mapdelete functions are already optimized to prefer direct
key comparison instead of tophash checks when key comparison is cheap.

Extended version of golang.org/cl/55235.

AMD64:
name                old time/op    new time/op    delta
MapPopulate/1         42.5ns ± 2%    40.3ns ± 2%  -5.37%  (p=0.000 n=9+10)
MapPopulate/10         558ns ± 1%     556ns ± 1%    ~     (p=0.157 n=10+10)
MapPopulate/100       7.75µs ± 1%    7.66µs ± 2%  -1.19%  (p=0.001 n=10+10)
MapPopulate/1000      92.6µs ± 1%    92.0µs ± 1%  -0.61%  (p=0.016 n=10+8)
MapPopulate/10000      817µs ± 1%     814µs ± 1%    ~     (p=0.247 n=10+10)
MapPopulate/100000    8.02ms ± 1%    7.90ms ± 2%  -1.47%  (p=0.007 n=10+10)

Change-Id: If0eca9931379cbbd37eb753e9bcd2888d8272153
Reviewed-on: https://go-review.googlesource.com/62050
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: revert "more compact representation of DW_AT_high_pc"
Heschi Kreinick [Thu, 7 Sep 2017 02:05:24 +0000 (22:05 -0400)]
cmd/compile: revert "more compact representation of DW_AT_high_pc"

This reverts commit 84188296e93dd4d26cb0a75a03a9096794e01e2f AKA CL 60530.

Fixes #21783

Change-Id: I68038a77de7446dea68419a40dd25982ea6d7df5
Reviewed-on: https://go-review.googlesource.com/62151
Reviewed-by: Heschi Kreinick <heschi@google.com>
7 years agodoc: update DWARF version
Hana Kim [Wed, 6 Sep 2017 20:17:11 +0000 (16:17 -0400)]
doc: update DWARF version

Change-Id: I0d02e5e44053f02fee3d84c782995ea4ce209a80
Reviewed-on: https://go-review.googlesource.com/62070
Reviewed-by: Heschi Kreinick <heschi@google.com>
7 years agocmd/compile: simplify exporting universal 'error' type
Matthew Dempsky [Wed, 6 Sep 2017 01:49:36 +0000 (18:49 -0700)]
cmd/compile: simplify exporting universal 'error' type

There shouldn't be any problems setting error's "Orig" (underlying)
type to a separate anonymous interface, as this is already how
go/types defines it.

Change-Id: I44e9c4048ffe362ce329e8306632e38b5ccfecff
Reviewed-on: https://go-review.googlesource.com/61790
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Alan Donovan <adonovan@google.com>
7 years agotesting: use time.Since instead of time.Now().Sub
Daniel Martí [Wed, 6 Sep 2017 19:13:06 +0000 (21:13 +0200)]
testing: use time.Since instead of time.Now().Sub

Change-Id: Ia0479f65a6ef033a65acf9ac8692777839a53b8c
Reviewed-on: https://go-review.googlesource.com/61990
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/internal/obj/x86: add some more AVX2 instructions
Agniva De Sarker [Tue, 5 Sep 2017 18:40:50 +0000 (00:10 +0530)]
cmd/internal/obj/x86: add some more AVX2 instructions

This adds the VFMADD[213|231]SD, VFNMADD[213|231]SD,
VADDSD, VSUBSD instructions

This will allow us to write a fast path for exp_amd64.s where
these optimizations can be applied in a lot of places.

Change-Id: Ide292107ab887bd1e225a1ad60880235b5ed7c61
Reviewed-on: https://go-review.googlesource.com/61810
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: more compact DWARF location for locals and arguments
Alessandro Arzilli [Tue, 4 Jul 2017 06:32:02 +0000 (08:32 +0200)]
cmd/compile: more compact DWARF location for locals and arguments

Now that all functions have a DW_AT_frame_base defined we can use
DW_OP_fbreg to specify the location of variables and formal parameters,
instead of the DW_OP_call_frame_cfa/DW_OP_consts/DW_OP_plus, saving 2
bytes for every variable and 2 bytes for every formal parameter after
the first one.

Change-Id: I2c7395b67e4a814a0131ab1520df11ca48ff9327
Reviewed-on: https://go-review.googlesource.com/60550
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
7 years agocmd/compile: more compact representation of DW_AT_high_pc
Alessandro Arzilli [Tue, 4 Jul 2017 07:11:04 +0000 (09:11 +0200)]
cmd/compile: more compact representation of DW_AT_high_pc

DWARF version 4 allows DW_AT_high_pc to be represented as a constant
offset from DW_AT_low_pc, this can help save up to 7 bytes per
function/lexical scope.

Change-Id: I93638d83638ecad4d0d1bfe27348eae6139820c9
Reviewed-on: https://go-review.googlesource.com/60530
Run-TryBot: Heschi Kreinick <heschi@google.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
7 years agocmd/asm: add most SSE4 missing instructions
isharipo [Mon, 21 Aug 2017 09:12:41 +0000 (12:12 +0300)]
cmd/asm: add most SSE4 missing instructions

Instructions added:
  INSERTPS immb, r/m, xmm
  MPSADBW immb, r/m, xmm
  BLENDPD immb, r/m, xmm
  BLENDPS immb, r/m, xmm
  DPPD immb, r/m, xmm
  DPPS immb, r/m, xmm
  MOVNTDQA r/m, xmm
  PACKUSDW r/m, xmm
  PBLENDW immb, r/m, xmm
  PCMPEQQ r/m, xmm
  PCMPGTQ r/m, xmm
  PCMPISTRI immb, r/m, xmm
  PCMPISTRM immb, r/m, xmm
  PMAXSB r/m, xmm
  PMAXSD r/m, xmm
  PMAXUD r/m, xmm
  PMAXUW r/m, xmm
  PMINSB r/m, xmm
  PMINSD r/m, xmm
  PMINUD r/m, xmm
  PMINUW r/m, xmm
  PTEST r/m, xmm
  PCMPESTRM immb, r/m, xmm

Note: only 'optab' table is extended.

`EXTRACTPS immb, xmm, r/m` is not included in this
change due to new ytab set 'yextractps'. This should simplify
code review.

4-operand instructions are a subject of upcoming changes that
make 4-th (and so on) operands explicit.
Related TODO note in asm6.go:
"dont't hide 4op, some version have xmm version".

Part of the mission to add missing amd64 SSE4 instructions to Go asm.

Change-Id: I71716df14a8a5332e866dd0f0d52d43d7714872f
Reviewed-on: https://go-review.googlesource.com/57470
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
7 years agocmd/asm: add amd64 PALIGNR instruction
isharipo [Fri, 18 Aug 2017 11:56:12 +0000 (14:56 +0300)]
cmd/asm: add amd64 PALIGNR instruction

3rd change out of 3 to cover AMD64 SSSE3 instruction set in Go asm.
This commit adds instruction that do require new ytab variable.

Change-Id: I0bc7d9401c9176eb3760c3d59494ef082e97af84
Reviewed-on: https://go-review.googlesource.com/56870
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/asm: add amd64 CLFLUSH instruction
isharipo [Fri, 18 Aug 2017 10:33:00 +0000 (13:33 +0300)]
cmd/asm: add amd64 CLFLUSH instruction

This is the last instruction I found missing in SSE2 set.

It does not reuse 'yprefetch' ytabs due to differences in
operands SRC/DST roles:
- PREFETCHx: ModRM:r/m(r) -> FROM
- CLFLUSH:   ModRM:r/m(w) -> TO

unaryDst map is extended accordingly.

Change-Id: I89e34ebb81cc0ee5f9ebbb1301bad417f7ee437f
Reviewed-on: https://go-review.googlesource.com/56833
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/asm: add amd64 PAB{SB,SD,SW}, PMADDUBSW, PMULHRSW, PSIG{NB,ND,NW}
isharipo [Fri, 18 Aug 2017 11:13:33 +0000 (14:13 +0300)]
cmd/asm: add amd64 PAB{SB,SD,SW}, PMADDUBSW, PMULHRSW, PSIG{NB,ND,NW}

instructions

1st change out of 3 to cover AMD64 SSSE3 instruction set in Go asm.
This commit adds instructions that do not require new named ytab sets.

Change-Id: I0c3dfd8d39c3daa8b7683ab163c63145626d042e
Reviewed-on: https://go-review.googlesource.com/56834
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocrypto/cipher: panic when IV length does not equal block size in NewOFB
Joe Kyo [Wed, 6 Sep 2017 09:31:03 +0000 (10:31 +0100)]
crypto/cipher: panic when IV length does not equal block size in NewOFB

Functions like NewCBCDecrypter, NewCBCEncrypter, NewCFBDecrypter,
NewCFBEncrypter and NewCTR all panic when IV length does not equal block size.
This commit changes NewOFB to panic too, instead of returning nil silently.

Change-Id: Ic4d3ebfad79bb0cf4759fa1c1a400c1a8d043490
Reviewed-on: https://go-review.googlesource.com/61850
Reviewed-by: Filippo Valsorda <hi@filippo.io>
Run-TryBot: Filippo Valsorda <hi@filippo.io>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: avoid stack allocation of a map bucket for large constant hints
Martin Möhrmann [Wed, 30 Aug 2017 23:00:39 +0000 (01:00 +0200)]
cmd/compile: avoid stack allocation of a map bucket for large constant hints

runtime.makemap will allocate map buckets on the heap for hints larger
than the number of elements a single map bucket can hold.

Do not allocate any map bucket on the stack if it is known at compile time
that hint is larger than the number of elements one map bucket can hold.
This avoids zeroing and reserving memory on the stack that will not be used.

Change-Id: I1a5ab853fb16f6a18d67674a77701bf0cf29b550
Reviewed-on: https://go-review.googlesource.com/60450
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: avoid redundant zeroing of hiter
Martin Möhrmann [Thu, 31 Aug 2017 19:26:03 +0000 (21:26 +0200)]
runtime: avoid redundant zeroing of hiter

The compiler and reflect already zero hiter before mapiterinit.

While here expand the documentation for mapiterinit.

Change-Id: I78b05d4d14bf78e8091e5353cdac80ffed30ca1e
Reviewed-on: https://go-review.googlesource.com/60673
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: set hiter type for map iterator in order pass
Martin Möhrmann [Sat, 2 Sep 2017 12:31:13 +0000 (14:31 +0200)]
cmd/compile: set hiter type for map iterator in order pass

Previously the type was first set to uint8 and then corrected
later in walkrange.

Change-Id: I9e4b597710e8a5fad39dde035df85676bc8d2874
Reviewed-on: https://go-review.googlesource.com/61032
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: move map ismapkey check to the compiler
Martin Möhrmann [Sat, 2 Sep 2017 15:32:58 +0000 (17:32 +0200)]
runtime: move map ismapkey check to the compiler

Remove the runtime ismapkey check from makemap and
add a check that the map key type supports comparison
to the hmap construction in the compiler.

Move the ismapkey check for the reflect code path
into reflect_makemap.

Change-Id: I718f79b0670c05b63ef31721e72408f59ec4ae86
Reviewed-on: https://go-review.googlesource.com/61035
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoflag: simplify arg logic in parseOne
alexpantyukhin [Tue, 5 Sep 2017 11:48:56 +0000 (15:48 +0400)]
flag: simplify arg logic in parseOne

Fixes #21763.

Change-Id: I59ee4f24c8064df64d9ede11aac02bc7ce4995b3
Reviewed-on: https://go-review.googlesource.com/61491
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoruntime: cleanup amd64p32 memmove and memclr file organization
Martin Möhrmann [Sat, 2 Sep 2017 06:33:21 +0000 (08:33 +0200)]
runtime: cleanup amd64p32 memmove and memclr file organization

Move memclr to a separate file to make it consistent
with other platforms asm function to file organization.

Remove nacl from the memmove filename as the implementation
is generic for the amd64p32 platform even if currently only
nacl is supported for amd64p32.

Change-Id: I8930b76da430a5cf2664801974e4f5185fc0f82f
Reviewed-on: https://go-review.googlesource.com/61031
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: remove global bout variable
Matthew Dempsky [Thu, 31 Aug 2017 00:43:19 +0000 (17:43 -0700)]
cmd/compile: remove global bout variable

Change-Id: I7054bbec080708c3a11ed62d7f6594e82aa33747
Reviewed-on: https://go-review.googlesource.com/61691
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
7 years agocmd/compile: fix evaluation order for OASOP
Matthew Dempsky [Tue, 29 Aug 2017 22:40:32 +0000 (15:40 -0700)]
cmd/compile: fix evaluation order for OASOP

Currently, we handle "x op= y" by rewriting as "x = x op y", while
ensuring that any calls or receive operations in 'x' are only
evaluated once. Notably, pointer indirection, indexing operations,
etc. are left alone as it's typically safe to re-evaluate those.

However, those operations were interleaved with evaluating 'y', which
could include function calls that might cause re-evaluation to yield
different memory addresses.

As a fix, simply ensure that we order side-effecting operations in 'y'
before either evaluation of 'x'.

Fixes #21687.

Change-Id: Ib14e77760fda9c828e394e8e362dc9e5319a84b2
Reviewed-on: https://go-review.googlesource.com/60091
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
7 years agocmd/compile: fix and improve struct field reflect information
Matthew Dempsky [Wed, 30 Aug 2017 21:17:24 +0000 (14:17 -0700)]
cmd/compile: fix and improve struct field reflect information

The previous logic was overly complicated, generated suboptimally
encoded struct type descriptors, and mishandled embeddings of
predeclared universal types.

Fixes #21122.
Fixes #21353.
Fixes #21696.
Fixes #21702.
Updates #21357.

Change-Id: If34761fa6dbe4af2af59dee501e7f30845320376
Reviewed-on: https://go-review.googlesource.com/60410
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years agocmd/compile: ignore non-code nodes when inlining
Mark Pulford [Sun, 3 Sep 2017 13:53:38 +0000 (23:53 +1000)]
cmd/compile: ignore non-code nodes when inlining

Avoid counting nodes that don't generate code (eg, constants) against the
inlining budget.

Fixes #21749

Change-Id: I10fca073e64be7d304709ef33e125eb8c78d5e4d
Reviewed-on: https://go-review.googlesource.com/61250
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: simplify "missing function body" error message
Matthew Dempsky [Sun, 3 Sep 2017 19:31:14 +0000 (12:31 -0700)]
cmd/compile: simplify "missing function body" error message

Fixes #21747.

Change-Id: I6a68370be3b7510ce364ddd1e41a1d767ce3a67f
Reviewed-on: https://go-review.googlesource.com/61311
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agodoc: Fixed missing dot in effective_go.html
Andrii Soldatenko [Sun, 3 Sep 2017 14:23:39 +0000 (17:23 +0300)]
doc: Fixed missing dot in effective_go.html

First sentence of https://golang.org/doc/effective_go.html#maps dot is missing.

Change-Id: Ic93b407945a6ad24a73ae9f7b6076ac9619871db
Reviewed-on: https://go-review.googlesource.com/61020
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
7 years agocmd/compile: fix assembly test
Keith Randall [Sun, 3 Sep 2017 15:52:34 +0000 (08:52 -0700)]
cmd/compile: fix assembly test

Bad merge, missed changing to keyed literal structs.

Bug introduced in CL 56252

Change-Id: I55cccff4990bd25e6387f6c90919ee5866900d7f
Reviewed-on: https://go-review.googlesource.com/61290
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: combine x*n - y*n into (x-y)*n
Cholerae Hu [Thu, 17 Aug 2017 07:06:42 +0000 (15:06 +0800)]
cmd/compile: combine x*n - y*n into (x-y)*n

Do the similar thing to CL 55143 to reduce IMUL.

Change-Id: I1bd38f618058e3cd74fac181f003610ea13f2294
Reviewed-on: https://go-review.googlesource.com/56252
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agomath: Add Round function (ties away from zero)
Mark Pulford [Thu, 18 May 2017 12:15:18 +0000 (22:15 +1000)]
math: Add Round function (ties away from zero)

This function avoids subtle faults found in many ad-hoc implementations,
and is simple enough to be inlined by the compiler.

Fixes #20100

Change-Id: Ib320254e9b1f1f798c6ef906b116f63bc29e8d08
Reviewed-on: https://go-review.googlesource.com/43652
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agoruntime: fix hashmap load factor computation
Keith Randall [Fri, 1 Sep 2017 19:32:38 +0000 (12:32 -0700)]
runtime: fix hashmap load factor computation

overLoadFactor wasn't really doing what it says it does.
It was reporting overOrEqualToLoadFactor.  That's actually what we
want when adding an entry to a map, but it isn't what we want when
constructing a map in the first place.

The impetus for this change is that if you make a map with a hint
of exactly 8 (which happens, for example, with the unitMap in
time/format.go), we allocate 2 buckets for it instead of 1.

Instead, make overLoadFactor really report when it is > the max
allowed load factor, not >=.  Adjust the callers who want to ensure
that the map is no more than the max load factor after an insertion
by adding a +1 to the current (pre-addition) size.

Change-Id: Ie8d85344800a9a870036b637b1031ddd9e4b93f9
Reviewed-on: https://go-review.googlesource.com/61053
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agonet/url: add examples to URL methods
Guilherme Garnier [Wed, 30 Aug 2017 23:06:23 +0000 (20:06 -0300)]
net/url: add examples to URL methods

Change-Id: Ia557771272a8ef3602f0281b8c89d2cfe146c8b3
Reviewed-on: https://go-review.googlesource.com/60470
Reviewed-by: Adrian Hesketh <adrianhesketh@hushmail.com>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile/internal/amd64: add MOVLloadidx8 and MOVLstoreidx8
Ilya Tocar [Tue, 20 Jun 2017 20:36:34 +0000 (15:36 -0500)]
cmd/compile/internal/amd64: add MOVLloadidx8 and MOVLstoreidx8

Currently we only use 1 and 4 as a scale for indexed 4-byte load.
In code generated in #20711 we can use indexed load with scale=8,
to improve performance:

name  old time/op  new time/op  delta
GM-6   108µs ± 0%    95µs ± 0%  -12.06%  (p=0.000 n=10+10)

So add new ops and combine loadidx1(shift 3..).. into loadidx8,
same for stores.

Change-Id: I5ed1c250ac40960e20606580cf9de221e75b72f1
Reviewed-on: https://go-review.googlesource.com/46134
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/go: pass plugin package name to compile -p
David Crawshaw [Fri, 1 Sep 2017 15:20:33 +0000 (11:20 -0400)]
cmd/go: pass plugin package name to compile -p

When compiling a plugin, package main gets a new name so as not to
conflict with the main package in the host binary, or any other
plugins. It is already defined by cmd/go, and used by cmd/link when
filling out the "" package placeholder in symbols.

With this CL, the plugin-specific name for main is also passed to
cmd/compile's -p flag. This is used to fill out the pkgpath field
of types, and ensures that two types defined in two different plugin
mains with the same name will not be mistaken for one another at
runtime.

Fixes #21386

Change-Id: I8a646d8d7451caff533fe0007343ea8b8e1704ed
Reviewed-on: https://go-review.googlesource.com/60910
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agospec: clarify context type for certain non-constant shifts
griesemer [Wed, 30 Aug 2017 13:10:12 +0000 (15:10 +0200)]
spec: clarify context type for certain non-constant shifts

The spec is not conclusive about whether a non-constant shift of
certain untyped constant left operands is valid when the shift
expression appears as an index in an index or slice expression,
or as a size in a `make` function call.

Despite identical spec rules in all these cases, cmd/compile accepts

make([]byte, 1.0 << s)

but pronounces an error for

a[1.0 << s]

(go/types accepts both).

This change clarifies the spec by explicitly stating that an
untyped constant left operand in a non-constant shift (1.0 in
the above examples) will be given type `int` in these contexts.

A separate issue #21693 addresses the cmd/compile bug.

Fixes #14844.

Change-Id: I4b52125e487a607fae377fcbed55463cdce9836c
Reviewed-on: https://go-review.googlesource.com/60230
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/tar: populate Devmajor and Devminor in FileInfoHeader on Darwin
Tobias Klauser [Thu, 31 Aug 2017 11:26:40 +0000 (13:26 +0200)]
archive/tar: populate Devmajor and Devminor in FileInfoHeader on Darwin

Extract device major/minor number on Darwin and set Devmajor and
Devminor in FileInfoHeader. Code based on the Major/Minor functions for
Darwin in golang.org/x/sys/unix.

Change-Id: I51b65f607bfa2e6b177b8b66e2b246b771367b84
Reviewed-on: https://go-review.googlesource.com/60850
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agogo/types: fix Info.Implicits entries
griesemer [Thu, 31 Aug 2017 15:41:51 +0000 (17:41 +0200)]
go/types: fix Info.Implicits entries

Packages of dot imports don't appear in the Info.Implicits map
since they are already taken care of by the Info.Defs map. Fix
documentation.

Implicitly dot-imported objects of a package shouldn't appear
in the Info.Implicits map because the documentation never said
so and there's no way to map multiple objects to the same
*ast.ImportSpec with the current data structure.

Added missing test for Info.Implicits.

The fix is a trivial one-line deletion, the rest is documentation
and test.

Fixes #21591.

Change-Id: I12a37dab85c531911c9363ec3d58daa095c7eb24
Reviewed-on: https://go-review.googlesource.com/60672
Reviewed-by: Alan Donovan <adonovan@google.com>
7 years agocmd/go: document the build flags properly for vet
Rob Pike [Fri, 1 Sep 2017 03:00:51 +0000 (13:00 +1000)]
cmd/go: document the build flags properly for vet

The "build flags" mentioned in the documentation are only those
that apply to analyzing packages and executing the tool.

Fixes #21711.

Change-Id: Ie7b2a354f1e30c928b40888c51fc68e599a5444a
Reviewed-on: https://go-review.googlesource.com/60830
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/cgo: support large unsigned macro again
Hiroshi Ioka [Thu, 31 Aug 2017 04:49:43 +0000 (13:49 +0900)]
cmd/cgo: support large unsigned macro again

The approach of https://golang.org/cl/43476 turned out incorrect.
The problem is that the sniff introduced by the CL only work for simple
expression. And when it fails it fallback to uint64, not int64, which
breaks backward compatibility.
In this CL, we use DWARF for guessing kind instead. That should be more
reliable than previous approach. And importanly, it fallbacks to int64 even
if it fails to guess kind.

Fixes #21708

Change-Id: I39a18cb2efbe4faa9becdcf53d5ac68dba180d46
Reviewed-on: https://go-review.googlesource.com/60510
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: fix clang option handling
Joel Sing [Thu, 31 Aug 2017 17:14:38 +0000 (03:14 +1000)]
cmd/go: fix clang option handling

On OpenBSD -current, clang is available/installed as 'cc'. This means that
the existing clang check fails and the clang related flags are not enabled.
Fix this by enabling the clang flags if the compiler claims to support them.

Change-Id: I84d124d2409a10f87002c6cbfdb69b4c9a55981a
Reviewed-on: https://go-review.googlesource.com/60750
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocrypto/cipher: extend the docs of BlockMode and Stream
Filippo Valsorda [Mon, 27 Feb 2017 12:45:02 +0000 (18:15 +0530)]
crypto/cipher: extend the docs of BlockMode and Stream

Change-Id: Iebb5b67c8defec22edd482d587edaf399a7ba82a
Reviewed-on: https://go-review.googlesource.com/37418
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agodoc/articles/wiki: fix final-test.patch
Matthew Dempsky [Thu, 31 Aug 2017 20:24:43 +0000 (13:24 -0700)]
doc/articles/wiki: fix final-test.patch

Fixes test regression introduced by golang.org/cl/53071.

Change-Id: Ibb085cf6d6be7b91b7c90b7bc6237c10f3097703
Reviewed-on: https://go-review.googlesource.com/60775
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
7 years agogo/types: escape +build sequence to silence vet warning
Matthew Dempsky [Thu, 31 Aug 2017 20:35:56 +0000 (13:35 -0700)]
go/types: escape +build sequence to silence vet warning

Appease the vet builder until we figure out the ideal solution.

Updates #21716.

Change-Id: I64ecdf9b6efb0933c5c79e11c860966ff1980305
Reviewed-on: https://go-review.googlesource.com/60776
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/link: emit DW_AT_data_member_location as a constant
Heschi Kreinick [Wed, 26 Jul 2017 19:58:25 +0000 (15:58 -0400)]
cmd/link: emit DW_AT_data_member_location as a constant

Simplify the DWARF representation of structs by emitting field offsets
as constants rather than location descriptions.

This was not explicitly mentioned as an option in DWARF2. It is
mentioned in DWARF4, but isn't listed in the changes, so it's not clear
if this was always intended to work or is an undocumented change. Either
way, it should be valid DWARF4.

Change-Id: Idf7fdd397a21c8f8745673ecc77ef65afa3ffe1c
Reviewed-on: https://go-review.googlesource.com/51611
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: David Chase <drchase@google.com>
7 years agoio/ioutil: don't cap buffer size in ReadFile
Bryan C. Mills [Tue, 15 Aug 2017 19:08:39 +0000 (15:08 -0400)]
io/ioutil: don't cap buffer size in ReadFile

When we added a Stat call to determine the initial buffer size in
https://golang.org/cl/163069, we included an arbitrary 1e9-byte limit
"just in case". That interacts badly with power-of-2 resizing in
*bytes.Buffer: it causes buffers reading from very large files to
consume up to twice the necessary space.

The documentation for (os.FileInfo).Size says that it reports "length
in bytes for regular files; system-dependent for others", but the
"system dependent" cases overwhelmingly return either a small number
(e.g., the length of the target path for a symlink) or a non-positive
number (e.g., for a file in /proc under Linux). It should be
appropriate to use the number reported by Size as an approximate lower
bound, even if it is large.

fixes #21455

Change-Id: I609c72519b7b87428c24d0b22db46eede30e0e54
Reviewed-on: https://go-review.googlesource.com/55870
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/fix: rewrite x/net/context by default
Sam Whited [Thu, 24 Aug 2017 17:39:34 +0000 (12:39 -0500)]
cmd/fix: rewrite x/net/context by default

Since these are type aliases in Go 1.9 make rewriting
golang.org/x/net/context imports to context the default.

Change-Id: Ib7ee58b59ea5463b92e1590c8b8f5c0ef87d310b
Reviewed-on: https://go-review.googlesource.com/58590
Reviewed-by: Sameer Ajmani <sameer@golang.org>
Run-TryBot: Sam Whited <sam@samwhited.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agodoc: add error handling on http.ListenAndServe
Francesc Campoy Flores [Thu, 3 Aug 2017 17:52:42 +0000 (10:52 -0700)]
doc: add error handling on http.ListenAndServe

Fixes #19511

Change-Id: I5585726773b822dba0be0196961132323ebbe084
Reviewed-on: https://go-review.googlesource.com/53071
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agonet/http: make startBackgroundRead panic if hijacked
Kunpei Sakai [Tue, 29 Aug 2017 09:56:59 +0000 (18:56 +0900)]
net/http: make startBackgroundRead panic if hijacked

Fixes #20933

Change-Id: I827c8d265674a7448c51c1da991e9b3b4501ba11
Reviewed-on: https://go-review.googlesource.com/59850
Run-TryBot: Tom Bergan <tombergan@google.com>
Reviewed-by: Tom Bergan <tombergan@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/internal/obj/arm: support more ARM VFP instructions
Ben Shi [Wed, 30 Aug 2017 08:20:43 +0000 (08:20 +0000)]
cmd/internal/obj/arm: support more ARM VFP instructions

Add support of more ARM VFP instructions in the assembler.
They were introduced in ARM VFPv2.

"NMULF/NMULD   Fm, Fn, Fd": Fd = -Fn*Fm
"MULAF/MULAD   Fm, Fn, Fd": Fd = Fd + Fn*Fm
"NMULAF/NMULAD Fm, Fn, Fd": Fd = -(Fd + Fn*Fm)
"MULSF/MULSD   Fm, Fn, Fd": Fd = Fd - Fn*Fm
"NMULSF/NMULSD Fm, Fn, Fd": Fd = -(Fd - Fn*Fm)

Change-Id: Icd302676ca44a9f5f153fce734225299403c4163
Reviewed-on: https://go-review.googlesource.com/60170
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agomath/big: fix internal comment
griesemer [Thu, 31 Aug 2017 13:02:24 +0000 (15:02 +0200)]
math/big: fix internal comment

Change-Id: Id003e2dbecad7b3c249a747f8b4032135dfbe34f
Reviewed-on: https://go-review.googlesource.com/60670
Reviewed-by: Marcel van Lohuizen <mpvl@golang.org>
7 years agogo/types: add +build ignore to file generated by hilbert test
griesemer [Thu, 31 Aug 2017 09:39:17 +0000 (11:39 +0200)]
go/types: add +build ignore to file generated by hilbert test

This makes sure that the go/types package still builds even if
the hilbert test generated its test file in the go/types package
(when run as: go test -run Hilbert -out=h.go).

Change-Id: I60ecbaaa1537de14cfa95e2e6fc8ebedff651baf
Reviewed-on: https://go-review.googlesource.com/60531
Reviewed-by: Marcel van Lohuizen <mpvl@golang.org>
7 years agocompress/flate: remove non-standard extensions to flate
Joe Tsai [Thu, 31 Aug 2017 01:49:18 +0000 (18:49 -0700)]
compress/flate: remove non-standard extensions to flate

Some constants were added to flate that seem to be an experimental
attempt at increasing the window size. However, according to RFC1951,
the largest window size is 32KiB, so these constants are non-standard.
Delete them.

Fixes #18458

Change-Id: Ia94989637ca031a56bce2548624fa48044caa7b9
Reviewed-on: https://go-review.googlesource.com/60490
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
7 years agoarchive/zip: support "end of central directory record comment"
Kenji Yano [Sat, 26 Aug 2017 09:44:27 +0000 (18:44 +0900)]
archive/zip: support "end of central directory record comment"

This change added support "end of central directory record comemnt" to the Writer.

There is a new exported field Writer.Comment in this change.
If invalid size of comment was set, Close returns error without closing resources.

Fixes #21634

Change-Id: Ifb62bc6c7f81b9257ac83eb570ad9915de727f8c
Reviewed-on: https://go-review.googlesource.com/59310
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: remove redundant constant shift rules
Cherry Zhang [Wed, 30 Aug 2017 16:11:29 +0000 (12:11 -0400)]
cmd/compile: remove redundant constant shift rules

Normal shift rules plus constant folding are enough to generate
efficient shift-by-constant instructions.

Add test to make sure we don't generate comparisons for constant
shifts.

TODO: there are still constant shift rules on PPC64. If they
are removed, the constant folding rules are not enough to remove
all the test and mask stuff for constant shifts. Leave them in
for now.

Fixes #20663.

Change-Id: I724cc324aa8607762d0c8aacf9bfa641bda5c2a1
Reviewed-on: https://go-review.googlesource.com/60330
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agonet/url: make Parse+String round trip magnet URLs
Johnny Luo [Sun, 16 Jul 2017 00:39:11 +0000 (10:39 +1000)]
net/url: make Parse+String round trip magnet URLs

Fixes #20054

Change-Id: I3c660ca0c56cdde2c2ac2f6a666d8531ab5588c5
Reviewed-on: https://go-review.googlesource.com/49050
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agoruntime: add symbols for Linux syscall numbers on 386/amd64
Chris Ball [Thu, 25 May 2017 19:48:02 +0000 (15:48 -0400)]
runtime: add symbols for Linux syscall numbers on 386/amd64

Matches other architectures by using names for syscalls instead of
numbers directly.

Fixes #20499.

Change-Id: I63d606b0b1fe6fb517fd994a7542a3f38d80dd54
Reviewed-on: https://go-review.googlesource.com/44213
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: fix regression in BenchmarkCompareBytes* for ppc64x
Carlos Eduardo Seo [Fri, 25 Aug 2017 18:00:00 +0000 (15:00 -0300)]
runtime: fix regression in BenchmarkCompareBytes* for ppc64x

Between go1.7 and go1.8, a performance regression was introduced in some of the
BenchmarkCompareBytes benchmarks.

Go1.7 vs Go1.8:
BenchmarkCompareBytesToNil-8               7.44          8.44          +13.44%
BenchmarkCompareBytesIdentical-8           6.96          11.5          +65.23%
BenchmarkCompareBytesBigIdentical-8        6.65          47112         +708351.13%

This change fixes the problem by optimizing the case where the byte slices being
compared are equal:

Go1.9 vs current:
BenchmarkCompareBytesToNil-8               7.35          7.00          -4.76%
BenchmarkCompareBytesIdentical-8           11.4          6.81          -40.26%
BenchmarkCompareBytesBigIdentical-8        48396         9.26          -99.98%

runtime.cmpstring can benefit from the same approach and is also changed.

Change-Id: I3cb25f59d8b940a83a2cf687eea764cfeff90688
Reviewed-on: https://go-review.googlesource.com/59650
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
7 years agocmd/cgo: support niladic function-like macros
Hiroshi Ioka [Tue, 23 May 2017 14:01:08 +0000 (23:01 +0900)]
cmd/cgo: support niladic function-like macros

Currently, cgo supports only macros which can be reduced to constants
or variables. The CL addresses remaining parts, macros which can be
represented as niladic functions.

The basic idea is simple:
  1. make a thin wrapper function per macros.
  2. replace macro expansions with function calls.

Fixes #10715
Fixes #18720

Change-Id: I150b4fb48e9dc4cc34466ef6417c04ac93d4bc1a
Reviewed-on: https://go-review.googlesource.com/43970
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/tar: minor doc fixes
Joe Tsai [Sat, 26 Aug 2017 09:12:56 +0000 (02:12 -0700)]
archive/tar: minor doc fixes

Use "file" consistently instead of "entry".

Change-Id: Ia81c9665d0d956adb78f7fa49de40cdb87fba000
Reviewed-on: https://go-review.googlesource.com/60150
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/tar: require opt-in to PAX or GNU format for time features
Joe Tsai [Fri, 25 Aug 2017 23:48:46 +0000 (16:48 -0700)]
archive/tar: require opt-in to PAX or GNU format for time features

Nearly every Header obtained from FileInfoHeader via the FS has
timestamps with sub-second resolution and the AccessTime
and ChangeTime fields populated. This forces the PAX format
to almost always be used, which has the following problems:
* PAX is still not as widely supported compared to USTAR
* The PAX headers will occupy at minimum 1KiB for every entry

The old behavior of tar Writer had no support for sub-second resolution
nor any support for AccessTime or ChangeTime, so had neither problem.
Instead the Writer would just truncate sub-second information and
ignore the AccessTime and ChangeTime fields.

In this CL, we preserve the behavior such that the *default* behavior
would output a USTAR header for most cases by truncating sub-second
time measurements and ignoring AccessTime and ChangeTime.
To use either of the features, users will need to explicitly specify
that the format is PAX or GNU.

The exact policy chosen is this:
* USTAR and GNU may still be chosen even if sub-second measurements
are present; they simply truncate the timestamp to the nearest second.
As before, PAX uses sub-second resolutions.
* If the Format is unspecified, then WriteHeader ignores AccessTime
and ChangeTime when using the USTAR format.

This ensures that USTAR may still be chosen for a vast majority of
file entries obtained through FileInfoHeader.

Updates #11171
Updates #17876

Change-Id: Icc5274d4245922924498fd79b8d3ae94d5717271
Reviewed-on: https://go-review.googlesource.com/59230
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime/internal/sys: use standard generated code header
Marvin Stenger [Wed, 30 Aug 2017 11:52:12 +0000 (13:52 +0200)]
runtime/internal/sys: use standard generated code header

This change implements the convention for generated code header agreed upon in https://golang.org/s/generatedcode.
Additionally run go generate.
Also update some comments.

Updates #13560

Change-Id: If45f91b93aaa0d43280c2c4630823bc4d2dc7d3a
Reviewed-on: https://go-review.googlesource.com/60250
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocontext: fix lint warning “drop = 0 from declaration”
Michael Stapelberg [Wed, 30 Aug 2017 14:50:13 +0000 (07:50 -0700)]
context: fix lint warning “drop = 0 from declaration”

Previously, the suggested code would result in the following golint warning:
“should drop = 0 from declaration of var errorsOnlyKey; it is the zero value”

Change-Id: I1a302c1e40ca89acbc76897e39097ecd04865460
Reviewed-on: https://go-review.googlesource.com/60290
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile/internal/ssa: remove redundant zeroextensions on amd64
Ilya Tocar [Wed, 23 Aug 2017 16:08:56 +0000 (11:08 -0500)]
cmd/compile/internal/ssa: remove redundant zeroextensions on amd64

Some instructions operating on <= 32 bits also zero out upper 32bits.
Remove zeroextensions of such values. Triggers a few times during
all.bash. Also removes ugly code like:
MOVL CX,CX

Change-Id: I66a46c190dd6929b7e3c52f3fe6b967768d00638
Reviewed-on: https://go-review.googlesource.com/58090
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agodoc: minor clarification regarding the sharing of underlying arrays
griesemer [Tue, 29 Aug 2017 13:48:07 +0000 (15:48 +0200)]
doc: minor clarification regarding the sharing of underlying arrays

The last sentence in the section on slice expressions could be read
as if it might apply to strings. Changed the sentence a bit to
emphasize its applicability to slices only. See also the issue for
more background.

Fixes #19220.

Change-Id: I8551f34230e4ed93f951e7b871cc81f54a5874a9
Reviewed-on: https://go-review.googlesource.com/59890
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: optimize ARM with MULS
Ben Shi [Fri, 25 Aug 2017 12:07:01 +0000 (12:07 +0000)]
cmd/compile: optimize ARM with MULS

MULS was introduced in ARMv7 and corresponding to MULA. This patch
duplicated all MULA related SSA rules with MULS.

Here was the contrast test result against the original go compiler.
There was no improvement in total, but big improvement in special cases.

1. A specific test case accelerated 18.62%.
(https://github.com/benshi001/ugo1/blob/master/mulsub_test.go)
name                     old time/op    new time/op    delta
MulSub-4                    270µs ± 0%     219µs ± 0%  -18.62%  (p=0.000 n=35+40)

2. Total size of all .a files in pkg/ shrank by 0.002%.

3. The compilecmp benchmark showed no decline.
name        old time/op       new time/op       delta
Template          2.37s ± 3%        2.36s ± 1%    ~     (p=0.233 n=19+18)
Unicode           1.32s ± 2%        1.34s ± 5%  +1.32%  (p=0.011 n=20+18)
GoTypes           7.88s ± 1%        7.87s ± 1%    ~     (p=0.758 n=20+20)
Compiler          37.5s ± 1%        37.6s ± 1%    ~     (p=0.194 n=20+19)
SSA               83.7s ± 2%        83.5s ± 2%    ~     (p=0.569 n=20+19)
Flate             1.46s ± 3%        1.45s ± 1%    ~     (p=0.619 n=20+17)
GoParser          1.87s ± 2%        1.85s ± 1%  -0.58%  (p=0.048 n=20+18)
Reflect           5.10s ± 2%        5.11s ± 2%    ~     (p=0.365 n=19+20)
Tar               1.78s ± 2%        1.78s ± 2%    ~     (p=0.531 n=19+20)
XML               2.62s ± 1%        2.61s ± 2%    ~     (p=0.057 n=17+19)
[Geo mean]        4.68s             4.67s       -0.07%

name        old user-time/op  new user-time/op  delta
Template          2.80s ± 1%        2.79s ± 2%    ~     (p=0.686 n=17+20)
Unicode           1.61s ± 4%        1.63s ± 6%    ~     (p=0.222 n=20+20)
GoTypes           9.59s ± 1%        9.60s ± 1%    ~     (p=0.482 n=17+20)
Compiler          46.1s ± 1%        46.2s ± 1%    ~     (p=0.373 n=20+18)
SSA                108s ± 1%         108s ± 2%    ~     (p=0.784 n=20+20)
Flate             1.68s ± 3%        1.69s ± 3%    ~     (p=0.335 n=20+19)
GoParser          2.20s ± 4%        2.19s ± 2%    ~     (p=0.844 n=20+18)
Reflect           5.97s ± 3%        6.01s ± 2%    ~     (p=0.184 n=20+20)
Tar               2.11s ± 2%        2.11s ± 4%    ~     (p=0.961 n=19+20)
XML               3.07s ± 1%        3.07s ± 3%    ~     (p=0.786 n=16+19)
[Geo mean]        5.61s             5.62s       +0.19%

name        old text-bytes    new text-bytes    delta
HelloSize         586kB ± 0%        586kB ± 0%    ~     (all equal)

name        old data-bytes    new data-bytes    delta
HelloSize        5.46kB ± 0%       5.46kB ± 0%    ~     (all equal)

name        old bss-bytes     new bss-bytes     delta
HelloSize        72.9kB ± 0%       72.9kB ± 0%    ~     (all equal)

name        old exe-bytes     new exe-bytes     delta
HelloSize        1.03MB ± 0%       1.03MB ± 0%    ~     (all equal)

4. The go1 benchmark showed no decline in total.
name                     old time/op    new time/op    delta
BinaryTree17-4              41.7s ± 1%     41.7s ± 1%    ~     (p=0.966 n=40+40)
Fannkuch11-4                23.6s ± 0%     23.6s ± 1%  -0.23%  (p=0.000 n=40+40)
FmtFprintfEmpty-4           844ns ± 1%     834ns ± 1%  -1.23%  (p=0.000 n=40+40)
FmtFprintfString-4         1.39µs ± 1%    1.40µs ± 1%  +0.71%  (p=0.000 n=40+40)
FmtFprintfInt-4            1.44µs ± 1%    1.45µs ± 1%  +0.70%  (p=0.000 n=40+40)
FmtFprintfIntInt-4         2.10µs ± 1%    2.10µs ± 1%  +0.30%  (p=0.000 n=40+40)
FmtFprintfPrefixedInt-4    2.49µs ± 0%    2.50µs ± 1%  +0.66%  (p=0.000 n=32+40)
FmtFprintfFloat-4          4.42µs ± 1%    4.46µs ± 2%  +0.94%  (p=0.000 n=40+40)
FmtManyArgs-4              8.31µs ± 1%    8.22µs ± 1%  -1.09%  (p=0.000 n=40+40)
GobDecode-4                 105ms ± 1%     102ms ± 1%  -2.30%  (p=0.000 n=39+39)
GobEncode-4                90.2ms ± 1%    88.7ms ± 1%  -1.66%  (p=0.000 n=40+39)
Gzip-4                      4.17s ± 1%     4.16s ± 1%    ~     (p=0.785 n=40+40)
Gunzip-4                    608ms ± 1%     608ms ± 1%    ~     (p=0.481 n=40+40)
HTTPClientServer-4          697µs ± 2%     684µs ± 3%  -1.89%  (p=0.000 n=37+40)
JSONEncode-4                255ms ± 1%     256ms ± 1%  +0.35%  (p=0.000 n=40+40)
JSONDecode-4                920ms ± 1%     926ms ± 1%  +0.64%  (p=0.000 n=40+39)
Mandelbrot200-4            49.3ms ± 1%    49.3ms ± 0%  +0.07%  (p=0.005 n=40+40)
GoParse-4                  46.8ms ± 2%    46.7ms ± 1%    ~     (p=1.000 n=40+40)
RegexpMatchEasy0_32-4      1.27µs ± 0%    1.27µs ± 1%    ~     (p=0.057 n=40+40)
RegexpMatchEasy0_1K-4      7.97µs ± 7%    7.92µs ± 5%    ~     (p=0.094 n=40+40)
RegexpMatchEasy1_32-4      1.28µs ± 1%    1.28µs ± 1%    ~     (p=0.406 n=40+40)
RegexpMatchEasy1_1K-4      10.5µs ± 4%    10.5µs ± 3%    ~     (p=0.855 n=40+40)
RegexpMatchMedium_32-4     2.04µs ± 0%    2.04µs ± 1%  -0.22%  (p=0.000 n=39+40)
RegexpMatchMedium_1K-4      541µs ± 0%     540µs ± 1%  -0.25%  (p=0.000 n=40+38)
RegexpMatchHard_32-4       29.3µs ± 1%    29.3µs ± 0%    ~     (p=0.149 n=40+40)
RegexpMatchHard_1K-4        878µs ± 1%     880µs ± 0%  +0.14%  (p=0.005 n=36+35)
Revcomp-4                  81.8ms ± 2%    81.4ms ± 2%  -0.43%  (p=0.015 n=38+39)
Template-4                  1.05s ± 1%     1.05s ± 1%    ~     (p=0.302 n=40+35)
TimeParse-4                7.18µs ± 1%    7.26µs ± 1%  +1.05%  (p=0.000 n=40+36)
TimeFormat-4               13.1µs ± 1%    13.1µs ± 1%    ~     (p=0.698 n=37+40)
[Geo mean]                  733µs          732µs       -0.16%

name                     old speed      new speed      delta
GobDecode-4              7.34MB/s ± 1%  7.51MB/s ± 1%  +2.36%  (p=0.000 n=39+39)
GobEncode-4              8.51MB/s ± 1%  8.65MB/s ± 1%  +1.69%  (p=0.000 n=40+39)
Gzip-4                   4.66MB/s ± 1%  4.66MB/s ± 1%    ~     (p=0.783 n=40+40)
Gunzip-4                 31.9MB/s ± 1%  31.9MB/s ± 1%    ~     (p=0.466 n=40+40)
JSONEncode-4             7.61MB/s ± 1%  7.58MB/s ± 1%  -0.35%  (p=0.001 n=40+40)
JSONDecode-4             2.11MB/s ± 1%  2.10MB/s ± 1%  -0.52%  (p=0.000 n=38+39)
GoParse-4                1.24MB/s ± 2%  1.24MB/s ± 1%    ~     (p=0.556 n=40+39)
RegexpMatchEasy0_32-4    25.1MB/s ± 0%  25.1MB/s ± 1%    ~     (p=0.064 n=40+40)
RegexpMatchEasy0_1K-4     129MB/s ± 8%   129MB/s ± 5%    ~     (p=0.094 n=40+40)
RegexpMatchEasy1_32-4    25.0MB/s ± 1%  25.1MB/s ± 1%    ~     (p=0.331 n=40+40)
RegexpMatchEasy1_1K-4    97.7MB/s ± 4%  97.8MB/s ± 3%    ~     (p=0.851 n=40+40)
RegexpMatchMedium_32-4    490kB/s ± 0%   490kB/s ± 0%    ~     (all equal)
RegexpMatchMedium_1K-4   1.89MB/s ± 0%  1.90MB/s ± 1%  +0.12%  (p=0.031 n=40+40)
RegexpMatchHard_32-4     1.09MB/s ± 1%  1.09MB/s ± 1%    ~     (p=0.597 n=40+40)
RegexpMatchHard_1K-4     1.16MB/s ± 1%  1.16MB/s ± 1%    ~     (p=0.565 n=40+35)
Revcomp-4                31.1MB/s ± 2%  31.2MB/s ± 2%  +0.44%  (p=0.018 n=38+39)
Template-4               1.85MB/s ± 1%  1.85MB/s ± 1%    ~     (p=0.873 n=40+40)
[Geo mean]               6.66MB/s       6.67MB/s       +0.26%

Change-Id: Icc972d8a78ea06c32c3aa15733ff0537c82c2dc7
Reviewed-on: https://go-review.googlesource.com/58950
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>

7 years agocmd/compile: rename mapbucket to bmap
Martin Möhrmann [Tue, 29 Aug 2017 19:10:17 +0000 (21:10 +0200)]
cmd/compile: rename mapbucket to bmap

This makes the name of the function to construct the map bucket type
consistent with runtimes naming and the existing hmap function.

Change-Id: If4d8b4a54c92ab914d4adcb96022b48d8b5db631
Reviewed-on: https://go-review.googlesource.com/59915
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: avoid padding for indirect map key or value on amd64p32
Martin Möhrmann [Tue, 29 Aug 2017 19:16:28 +0000 (21:16 +0200)]
cmd/compile: avoid padding for indirect map key or value on amd64p32

Padding needed for map buckets is dependent on the types used to
construct the map bucket. In case of indirect keys or values pointers
are used in the map bucket to the keys or values.

Change the map bucket padding calculation to take the alignment of
the key and value types used to construct the map bucket into account
instead of the original key and value type.

Since pointers are always 32bit aligned on amd64p32 this prevents
adding unneeded padding in case the key or value would have needed
64bit alignment without indirect referencing.

Change-Id: I7943448e882d269b5cff7e921a2a6f3430c50878
Reviewed-on: https://go-review.googlesource.com/60030
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: simplify check for pointers in map bucket
Martin Möhrmann [Tue, 29 Aug 2017 19:32:41 +0000 (21:32 +0200)]
cmd/compile: simplify check for pointers in map bucket

Change-Id: Ie5e977c54f50274421962a7ca268ae71a6f19cef
Reviewed-on: https://go-review.googlesource.com/60050
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: move dynamic makemap checks into cmd/compile
Martin Möhrmann [Tue, 29 Aug 2017 11:31:45 +0000 (13:31 +0200)]
runtime: move dynamic makemap checks into cmd/compile

Check map invariants, type size and alignments during compile time.

Keep runtime checks for reflect by adding them to reflect_makemap.

Change-Id: Ia28610626591bf7fafb7d5a1ca318da272e54879
Reviewed-on: https://go-review.googlesource.com/59914
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/link: unexport all peSection fields
Alex Brainman [Tue, 29 Aug 2017 03:30:17 +0000 (13:30 +1000)]
cmd/link: unexport all peSection fields

Change-Id: I83e168f0d1dd1897a0c02c0f1233e1054e93fb0f
Reviewed-on: https://go-review.googlesource.com/59791
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.addDWARF
Alex Brainman [Tue, 29 Aug 2017 03:09:59 +0000 (13:09 +1000)]
cmd/link: introduce and use peFile.addDWARF

Change-Id: I2c217e03779772605aa0b5a33ef80459333eeebc
Reviewed-on: https://go-review.googlesource.com/59790
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: refactor container()
Hiroshi Ioka [Mon, 28 Aug 2017 06:00:49 +0000 (15:00 +0900)]
cmd/link: refactor container()

* rename to emitPcln because I'd like to skip not only container types,
  but also something like "go.buildid" in the future.
* return bool instead of int.

Change-Id: I029adb81292f7dd2fe98e69f3877c5c27f32ec30
Reviewed-on: https://go-review.googlesource.com/59415
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

7 years agoarchive/tar: populate uname/gname/devmajor/devminor in FileInfoHeader
Joe Tsai [Mon, 28 Aug 2017 23:25:56 +0000 (16:25 -0700)]
archive/tar: populate uname/gname/devmajor/devminor in FileInfoHeader

We take a best-effort approach since information for these fields
are not well supported on all platforms.

user.LookupId+user.LookupGroupId is currently 15x slower than os.Stat.
For performance reasons, we perpetually cache username and groupname
with a sync.Map. As a result, this function will not be updated whenever
the user or group names are renamed in the OS. However, this is a better
situation than before, where those fields were not populated at all.

Change-Id: I3cec8291aed7675dea89ee1cbda92bd493c8831f
Reviewed-on: https://go-review.googlesource.com/59531
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>